From 6662d65adfb9133e251a098b8563a62f8236e637 Mon Sep 17 00:00:00 2001 From: Danny Garside Date: Wed, 11 Dec 2019 23:26:47 +0000 Subject: [PATCH] post-corrections, checking stage --- Conclusions.tex | 8 +- Interviews.tex | 6 +- LargeSphere.tex | 57 +++++++++---- LitReview.tex | 10 +++ Main.tex | 12 +-- MathMethods.tex | 5 +- MethodsForCC.tex | 1 + SmallSphere.tex | 80 +++++++++++++------ figs/SmallSphere/LED_SPDcontrast.pdf | Bin 0 -> 20135 bytes figs/SmallSphere/SSseccal2.pdf | Bin 0 -> 19626 bytes figs/SmallSphere/ScreenShieldingAnalysis.pdf | Bin 0 -> 9369 bytes ipRGCs.tex | 1 + melcomp.tex | 6 +- 13 files changed, 133 insertions(+), 53 deletions(-) create mode 100644 figs/SmallSphere/LED_SPDcontrast.pdf create mode 100644 figs/SmallSphere/SSseccal2.pdf create mode 100644 figs/SmallSphere/ScreenShieldingAnalysis.pdf diff --git a/Conclusions.tex b/Conclusions.tex index c4746d3..7a334a3 100644 --- a/Conclusions.tex +++ b/Conclusions.tex @@ -6,10 +6,10 @@ \section{Summary of Conclusions} \begin{enumerate} \item Assuming the applicability of general damage functions, potential damage may be reduced by using light sources with lower \glspl{CCT} (Figure \ref{fig:CCTvsDI}). \item Museum professionals currently do not employ this technique, in part due to conflicting results as to how \gls{CCT} interacts with observer preference (Section \ref{sec:CCTmus}). -\item One potential cause for these conflicting results (involvement of melanopsin in colour constancy) was examined with results showing no evidence for a strong or simple interaction (Chapters \ref{chap:LargeSphere} and \ref{chap:SmallSphere}). -\item To address limitations in the performed experiments, and assist with designing future experiments, a computational study was performed to explore whether a melanopsin interaction would be beneficial for colour constancy (Chapter \ref{chap:Melcomp}). +\item One potential cause for the conflicting results of experiments looking to find a canonical preferred \gls{CCT} might be if \glspl{ipRGC} were involved in colour constancy or chromatic adaptation. This possibility was explored with two psychophysical experiments. The results showed no evidence for a strong or simple interaction (Chapters \ref{chap:LargeSphere} and \ref{chap:SmallSphere}). +\item To address limitations in the performed experiments, and assist with designing future experiments, a computational study was performed to explore whether a melanopsin interaction would be beneficial for colour constancy (Chapter \ref{chap:Melcomp}). It was found that there was value in a melanopic input to chromatic adaptation or colour constancy. It was also found that the spectral sensitivity of melanopsin is close to optimal for providing this type of signal. \item In response to this, and observations from psychophysical work, future work is proposed (Section \ref{sec:fut}). -\item A novel method for performing colour constancy experiments with a tablet computer has been provided, and recommendations for further development are suggested (Chapter \ref{chap:Tablet}). +\item A novel method for performing colour constancy experiments with a tablet computer has been provided, and recommendations for further development are suggested (Chapter \ref{chap:Tablet}). This method seems valuable since it allows for colour constancy experiments to be performed in natural environments. It was found that overall this methodology is effective, with certain caveats and controls required. \end{enumerate} \section{Contributions to Museum Lighting} @@ -28,6 +28,8 @@ \section{Contributions to Vision Science} However, since based on the results of \citet{kraft_mechanisms_1999} it seems likely that colour constancy is achieved through the interplay of a great many factors, and there is not a single panacea to the problem of illuminant variance, special care should be taken. It is likely that in some situations certain mechanisms or cues are given weight, where in another situation, others are prioritised. +Since at this time there is no clear indication of the level at which the visual system might use a melanopic signal, it seems wise to focus future experiments on high-level and naturalistic behavioural tasks such as colour naming and differentiation of surface from illumination changes, rather than techniques which potentially primarily probe low level mechanisms, such as achromatic matching. + \section{Future work} \label{sec:fut} Future work relevant to the computational study and the tablet method is listed at the end of the respective chapters. A number of recommendations for future sphere-type psychophysical investigations are provided below. diff --git a/Interviews.tex b/Interviews.tex index 88337dd..e751452 100644 --- a/Interviews.tex +++ b/Interviews.tex @@ -83,4 +83,8 @@ \subsection{Colour Rendering} \section{Conclusion} Generally, the interviewees believed that visitor requirements were being met (although there is often difficulty in defining exactly what visitor requirements actually are) and no specific tool or technology was proposed that would provide a clear benefit. Several interviewees mentioned that a way to improve the accessibility of colour rendering indices (in terms of the ease with which they could be understood and applied) would be appreciated. No interviewees knew of recent surveys similar to the present one. -The impact that these interviews had on the research which followed is such; it became clear that \gls{CCT} was a tool which could be used to reduce damage, but which was not being used at the time. One of the barriers to use was a lack of understanding of how \gls{CCT} interacted with other visual appearance properties and preference. It therefore seemed valuable to attempt to extend our understanding of chromatic adaptation and colour constancy, with the hope that this would allow museum professionals to limit damage through specification of lower values of \gls{CCT}. \ No newline at end of file +\section{Interim Summary} + +The impact that these interviews had on the research which followed is such; it became clear that \gls{CCT} was a tool which could be used to reduce damage, but which was not being used at the time. One of the barriers to use was a lack of understanding of how \gls{CCT} interacted with other visual appearance properties and preference. It therefore seemed valuable to attempt to extend our understanding of chromatic adaptation and colour constancy, with the hope that this would allow museum professionals to limit damage through specification of lower values of \gls{CCT}. + +The following chapters all seek to extend our knowledge of colour constancy and chromatic adaptation. \ No newline at end of file diff --git a/LargeSphere.tex b/LargeSphere.tex index 06a02d0..a69feb8 100644 --- a/LargeSphere.tex +++ b/LargeSphere.tex @@ -31,7 +31,7 @@ \subsection{Hardware} The interior of the sphere was painted with RAL 7040 dulux vinyl matt grey, of approximately 38\% reflectance. -Illumination was provided by a Kodak slide projector with a tungsten-halogen light source, filtered through one of 16 near-monochromatic filters, ranging in 20nm intervals from 400-700nm inclusive. Measurements of the internal illumination were taken with a \gls{PR650} device, plotted in Figure \ref{fig:LSillum}. The illumination has been described further in \citet{macdonald_chromatic_2013}: ``The average luminance of the surrounding chromatic adapting field ranged from a maximum of 0.75 cd/m2 at 560 nm to less than 0.05 cd/m2 at the ends of the spectrum, corresponding to a retinal illuminance through a pupil of diameter 8mm ranging from 38 trolands (max) to less than 2.5 trolands, meaning that the viewing environment was in the upper mesopic range''. No effort was made to ensure that light from inside the sphere did not hit and/or reflect from the LCD display. +Illumination was provided by a Kodak slide projector with a tungsten-halogen light source, filtered through one of 16 near-monochromatic filters, ranging in 20nm intervals from 400-700nm inclusive (Figure \ref{fig:LSillum}). During each session, only one filter was used and the surround illumination remained the same throughout. Measurements of the internal illumination were taken with a \gls{PR650} device, plotted in Figure \ref{fig:LSillum}. The illumination has been described further in \citet{macdonald_chromatic_2013}: ``The average luminance of the surrounding chromatic adapting field ranged from a maximum of 0.75 cd/m2 at 560 nm to less than 0.05 cd/m2 at the ends of the spectrum, corresponding to a retinal illuminance through a pupil of diameter 8mm ranging from 38 trolands (max) to less than 2.5 trolands, meaning that the viewing environment was in the upper mesopic range''. No effort was made to ensure that light from inside the sphere did not hit and/or reflect from the LCD display. \begin{figure}[htbp] \includegraphics[max width=\textwidth]{figs/LargeSphere/LSillum.pdf} @@ -41,7 +41,11 @@ \subsection{Hardware} \subsection{Observer task} -The observer sat on one side of the sphere with their face inside the sphere (as shown in Figure \ref{fig:Alejandro}), such that nothing outside of the sphere was visible. On view on the opposite side of the sphere was a circular 4$^{\circ}$ aperture onto an LCD screen, upon which a random colour was visible (See section \ref{sec:LSstim} for further details of the randomisation starting routine). It was the observer's task to use two handheld sliders, which controlled the chromaticity of the screen, to make the appearance of the screen achromatic (an achromatic setting task). On average it took observers roughly 20 seconds to make a selection. Once the observer was happy with the achromacy of the patch, a button was pressed to record the setting and a new random colour would be presented. The first displayed colour was at \gls{CIE} L* (of CIELAB and CIELUV) of 85, with subsequent colours descending by 5 L* until 10 L*. This sequence was repeated 10 times per session. Per session observers made 10 selections at 16 lightness levels (160 total). Observers performed 16 sessions (2560 selections total), one session for each adapting wavelength, visualised in Figure \ref{fig:ExperimentalPro}. Observers found sessions quite fatiguing and generally did not wish to do more than two or three sessions per day. A brief break was generally taken between sessions, though minimum time for such was prescribed. +The observer sat on one side of the sphere with their face inside the sphere (as shown in Figure \ref{fig:Alejandro}), such that nothing outside of the sphere was visible. On view on the opposite side of the sphere was a circular 4$^{\circ}$ aperture onto an LCD screen, upon which a random colour was visible (See section \ref{sec:LSstim} for further details of the randomisation starting routine). Surrounding the aperture, the rest of the ganzfeld was illuminated by light from the slide projector, filtered through one of 16 near-monochromatic filters (Figure \ref{fig:LSillum}). It was the observer's task to use two handheld sliders, which controlled the chromaticity of the screen, to make the appearance of the screen achromatic (an achromatic setting task). + +On average it took observers roughly 20 seconds to make a selection. Once the observer was happy with the achromacy of the patch, a button was pressed to record the setting and a new random colour would be presented. The first displayed colour was at \gls{CIE} L* (of CIELAB and CIELUV) of 85, with subsequent colours descending by 5 L* until 10 L*. + +This sequence was repeated 10 times per session. Per session observers made 10 selections at 16 lightness levels (160 total). Observers performed 16 sessions (2560 selections total), one session for each surround adapting wavelength. The overall protocol is visualised in Figure \ref{fig:ExperimentalPro}. Observers found sessions quite fatiguing and generally did not wish to do more than two or three sessions per day. A brief break was generally taken between sessions, though minimum time for such was prescribed. For one observer, in an additional (17th) session the narrow-band filter was replaced by a neutral density filter, to produce an achromatic adapting field. @@ -60,11 +64,11 @@ \subsection{Observer task} \subsection{Stimulus specification} \label{sec:LSstim} -The stimulus was controlled via a MATLAB script, which read the input of two sliders and a button via a `Phidget' interface. The two linear sliders provided values of between 0 and 1000, and these values were converted to approximate CIELAB co-ordinates in such a manner that the slider maxima corresponded to the Natural Colour System (NCS) unique hue positions as computed by \citet{derefeldt_transformation_1986}. In this manner the sliders could be considered as moving along opponent axes between red and green (via the CIELAB origin), and blue and yellow (via the CIELAB origin) respectively. These values were transformed into standard XYZ values, with white references of [XYZ = 99.04, 100, 151.30] for observer LM and [XYZ = 94.97, 100, 98.15] for observer TR. It is unclear why these values were chosen, though it is suspected that at least one of them was the native white point for the display as measured at the time. These values were then converted to display RGB signals using the GOG characterisation model and output to the screen. +The stimulus was controlled via a MATLAB script, which read the input of two sliders and a button via a `Phidget' interface. The two linear sliders provided values of between 0 and 1000, and these values were converted to approximate CIELAB co-ordinates in such a manner that the slider maxima corresponded to the Natural Colour System (NCS) unique hue positions as computed by \citet{derefeldt_transformation_1986}. In this manner the sliders could be considered as moving along opponent axes between red and green (via the CIELAB origin), and blue and yellow (via the CIELAB origin) respectively. -The generation of random starting colours was achieved by modulating the nominal zero-point on each slider scale, where the default zero point is considered as 500, sampling from a uniform distribution between 250 and 750 for each presentation. +These values were transformed into XYZ values, with white references of [XYZ = 99.04, 100, 151.30] for observer LM and [XYZ = 94.97, 100, 98.15] for observer TR. The white reference for LM related to a screen characterisation performed around the time that the initial measurements were made. It is assumed that the same is true of the white point used for observer TR, although this characterisation data is no longer available. These values were then converted to sRGB values, and output to screen. -The slider position was then considered relative to this new zero-point, meaning that once an observer hit the button to confirm their selection, a new random colour, anchored to their previous selection (in a rough manner) would be presented. It should be noted that this new colour was not entirely independent of the previous selection, due to this anchoring. +The generation of random starting colours was achieved by modulating the nominal zero-point on each slider scale, where the default zero point is considered as 500, sampling from a uniform distribution between 250 and 750 for each presentation. The slider position was then considered relative to this new zero-point. The effect of this can be considered as such: if the observer were to leave the slider in a central position throughout (make no selections) they would be provided with random colours drawn uniformly from within bounds either side of the objective white-point. Since, hopefully, the observer was making selections, each new `random' colour would be biased towards the previous selection (based on where the sliders had been left from the previous selection). It should be noted that this new colour was not entirely independent of the previous selection, due to this bias. % \subsection{Data Collection} @@ -74,17 +78,17 @@ \subsection{Data Processing} Data were calibrated in the following manner: the recorded RGB values of the observers' selections were bounded (values above 1 or below 0, which occurred when observers made selections which were outside of the sRGB gamut, were brought within range, with an `absolute' rendering intent), quantized to 8-bits, and converted via look-up-table to \gls{CIE} 1931 XYZ tristimulus values. From these, xy chromaticities and CIELAB values were computed (with the white point of the display, as loaded from the characterisation file, used as the white reference). -A set of data referred to as the `baseline' data was generated, where there was no observer input, to consider the range of possible responses that an observer could make. This data was processed in exactly the same manner as the observer data, and is shown in Figure \ref{fig:overviewBL}, where it can be seen that the chromatic gamut increases as L* increases (due to a factor \texttt{cfac} in the display code which aimed to scale chromatic space with L*, mirroring the shape of CIELAB). It can also be seen that the gamut boundary is sometimes reached at higher levels of L*, with some of the vectors curving to remain within gamut. +A set of data referred to as the `baseline' data was generated, where there was no observer input, to consider the range of possible responses that an observer could make. The element of code which provided new random starting positions was excluded for these sessions. This data was processed in exactly the same manner as the observer data, and is shown in Figure \ref{fig:overviewBL}, where it can be seen that the chromatic gamut increases as L* increases (due to a factor \texttt{cfac} in the display code which aimed to scale chromatic space with L*, mirroring the shape of CIELAB). It can also be seen that the gamut boundary is sometimes reached at higher levels of L*, with some of the vectors curving to remain within gamut. \begin{figure}[htbp] \includegraphics[max width=1.2\textwidth, center]{figs/LargeSphere/baselinedataOverview.pdf} -\caption{The baseline dataset. This data represents the condition where there is no observer input, and the two sliders are left at their maximum (1000), minimum (0), or neutral (500) positions (9 combinations). In the legend the first number denotes the slider A position, and the second denotes the slider B position.} +\caption{The baseline dataset. This data represents the condition where there is no observer input. The points on the curves are the different values of L* presented during a session. Each curve represents a single `session'. The two sliders are left at their maximum (1000), minimum (0), or neutral (500) positions (9 combinations). This was done computationally; no actual slider input was used. In the legend the first number denotes the slider A position, and the second denotes the slider B position.} \label{fig:overviewBL} \end{figure} %The effect of the random offsetting was also queried, and it was found that the maximum random offset pushed the chromaticities halfway between \dots -Two distinct approaches were taken to data analysis. The first attempted to process the data in a chromatic space, with the reasoning that under the null condition chromatic selections should simply correspond to the chromaticity of the surround illuminations, presumably with some sort of gain function applied. If it could be shown that this relationship was not as expected, in a manner which might suggest involvement by other mechanisms (meaning rods or \glspl{ipRGC}), then this could be considered as evidence against the null hypothesis. +Two distinct approaches were taken to data analysis. The first attempted to process the data in a chromatic space, with the reasoning that under the null condition chromatic selections should simply correspond to the chromaticity of the surround adapting illuminations, presumably with some sort of gain function applied. If it could be shown that this relationship was not as expected, in a manner which might suggest involvement by other mechanisms (meaning rods or \glspl{ipRGC}), then this could be considered as evidence against the null hypothesis. The second approach took advantage of the fact that measurements were taken at samples across the wavelength spectrum. Since we know the power at each waveband, and the spectral sensitivities of the receptors, we can see whether the responses (transformed to cone space) relate in some simple way to spectral sensitivities. @@ -124,7 +128,7 @@ \subsection{Variability over time/repeats} Not represented in the above plots is the way in which responses varied over time within each session, averaged over L* (the previous plots were averages over time). Figures \ref{fig:timeLM} and \ref{fig:timeTR} show the calibrated CIELAB values for observers LM and TR. -L* should remain steady throughout (this was set), with minor differences introduced presumably due to differences between the screen and sRGB, 8-bit quantization, and any selections where a gamut boundary was reached. Both a* and b* follow the broad trends which would be expected given previous figures. Newly visible in these figures is the manner in which responses change over time. For both observers LM and TR the first two or three repeats seem to be distinct from the rest of the set, suggesting that adaptation had not yet reached a steady-state during this time. +L* should remain steady throughout (this was set), with minor differences introduced presumably due to differences between the screen and sRGB, 8-bit quantization, and any selections where a gamut boundary was reached. Both a* and b* follow the broad trends which would be expected given previous figures. Newly visible in these figures is the manner in which responses change over time. For both observers LM and TR the first two or three repeats seem to be distinct from the rest of the set, suggesting that adaptation had not yet reached a steady-state during this time. Further quantitative analysis is provided in the following section. \begin{figure}[htbp] \includegraphics[max width=1.2\textwidth, center]{figs/LargeSphere/LMdataOverTime.pdf} @@ -148,7 +152,7 @@ \subsection{Chromaticity-based analysis} \begin{figure}[htbp] \includegraphics[max width=\textwidth]{figs/LargeSphere/adapter1.pdf} -\caption{The CIELAB values for the surround illuminations, calculated from the measurements shown in Figure \ref{fig:LSillum}, taking the white point of the screen (for the TR trials) as the white point.} +\caption{The CIELAB values for the surround adapting illuminations, calculated from the measurements shown in Figure \ref{fig:LSillum}, taking the white point of the screen (for the TR trials) as the white point.} \label{fig:adapter1} \end{figure} @@ -162,7 +166,7 @@ \subsection{Chromaticity-based analysis} \begin{figure}[htbp] \includegraphics[max width=\textwidth]{figs/LargeSphere/TRcompareWithSurround.pdf} -\caption{As per Figure \ref{fig:LMCompSurr} but for the data of TR.} +\caption{As per Figure \ref{fig:LMCompSurr} but for the data of TR. Note that the white point used for visualisation relate to the white points used during data collection, which differ for each observer.} \label{fig:TRCompSurr} \end{figure} @@ -183,6 +187,8 @@ \subsubsection{Colour Constancy Indices} where $b$ is the distance between the post-adaptation point and the ideal match, and $a$ is the distance between the pre-adaptation point and the ideal match\footnote{For further discussion see \citet[Section 4.1, pg. 681]{foster_color_2011}.}. +There are multiple reasonable options for which value to use as a `pre-adaptation point'. First, the origin of the space within which selections are made (different for each observer) seems to be a possible option; this corresponds to the central point on each slider over time for each individual. However, though the set-up ascribes some value to this point, it is not definitively linked to the settings that observers made; it can be seen in Figures \ref{fig:LMCompSurr} and \ref{fig:TRCompSurr} that there seems to be no particular relevance of the point [0,0]. A second option would be to use the measurements made under a neutral density filter for both observers. This data has not been used thus far. However, again there is no actual significance of these values - a neutral density filter could be slightly chromatic and still be labelled as a neutral density filter, and even if it were perfectly spectrally neutral, it's designation as a gold standard `neutral' only actually passes on responsibility to the chromaticity of the projector lamp, which is under no obligation to be especially `neutral'. The third option is to use the average setting value, which has no specific logical background, but is vastly more practically relevent than the previous two options. This third option was chosen for future analyses. + Averaging over time for each observer, and using the average response for each observer as the pre-adaptation point yields \glspl{CCI} as shown in Figures \ref{fig:LMCCI} and \ref{fig:TRCCI}. Only data for L* of 20 and 60 is plotted for clarity, in keeping with previous figures. It can be seen that there are common trends across wavelength at the different values of L*. \begin{figure}[htbp] @@ -235,7 +241,7 @@ \subsubsection{Colour Constancy Indices} \label{fig:TRCCI_T} \end{figure} -A multi-way ANOVA performed upon the data, treating wavelength, time and L* and independent categorical variables found a significant effect of each, as shown in Figures \ref{fig:anova} and \ref{fig:anova}, with a level of alpha of 0.05. +A three-way ANOVA performed upon the data, treating wavelength, time and L* and independent categorical variables found a significant effect of each, as shown in Figures \ref{fig:anova} and \ref{fig:anova2}, with a level of $\alpha$ of 0.05. \begin{figure}[htbp] \includegraphics[max width=\textwidth]{figs/LargeSphere/anova.png} @@ -249,7 +255,11 @@ \subsubsection{Colour Constancy Indices} \label{fig:anova2} \end{figure} -\clearpage +Whilst variables were treated as categorical in the above analysis, there would be an argument for treating each as a continuous variable. However, several factors would need to be considered. Foremost, whilst wavelength is nominally a continuous variable, in this experiment each wavelength category had a difference level of radiance, which means that caution should be taken in assuming their equivalence. It is also possible that each filter might have a meaningfully different spectral transmission profile, specifically the band-pass width (see Figure \ref{fig:LSillum}). + +Caution is also required regarding the assumption of independence of measurements. Due to the nature of chromatic adaptation, it would not be possible to interleave conditions, and so wavelength is further confounded with various other factors; date, time of day, and all manner of secondary factors relating to these (whether the observer has eaten recently for example). + +It would be of interest to assess whether the contrast between the surround and the selection area influenced settings, but for this specific experiment the contrast is confounded by wavelength and L* and there does not seem to be a clear way to examine the influence of contrast directly. % But luminance \dots @@ -261,9 +271,18 @@ \subsubsection{Colour Constancy Indices} \subsection{Spectrum-based analysis} -%\textit{The code to reproduce the following analysis can be found at \url{https://github.com/da5nsy/LargeSphere/blob/master/Data\%20Analysis/VonKriesTest.m}} +This analysis aimed to leverage the fact that we have access to estimates of the of the spectral emission of the screen for each RGB value, and thus can calculate the relative cone/rod/\gls{ipRGC} catches for each achromatic setting. This in turn allows us to test to what extent the results seen can be explained simply by scaling cone mechanisms (a \emph{diagonal}, or Von-Kries-type transformation) and to ask whether adding additional inputs to the model (rods and/or \glspl{ipRGC}) improves our ability to explain the measured results. + +The first stage of this analysis was to generate simulated data which represented the situation whereby there was only simple Von Kries adaptation. Under this situation the estimated cone catches for the achromatic matches would be equal to the \glspl{SSF} of the cones, linearly scaled by the radiance levels of each surround adapting field. If the radiance levels were equal at each wavelength interval, the simulated data would be equal to the cone \glspl{SSF}. If, hypothetically, one wavelength interval were vastly higher in radiance, we would expect a correspondingly higher adaptive effect, which would result in a higher level of activation required in order for an achromatic visual appearance. In this way, we predict what results may look like if the only type of adaptation occurring was a simple Von Kries / diagonal scaling. -The first stage of this analysis was to generate simulated data which represented the situation whereby there was only simple Von Kries adaptation. This was accomplished by multiplying the individual \glspl{SPD} (Figure \ref{fig:LSillum}) by the CIE 2006 10$^{\circ}$ observer fundamentals (not in the manner in which this is usually done, resulting in tristimulus values, but rather point-wise, thus retaining the spectral nature of the data). See Figure \ref{fig:LSsimdata}. +Practically, this is accomplished by element-wise multiplication of each sensor \gls{SSF} by the measured emission from each adapting surround, as per Equation \ref{eq:VK}. The absolute scaling, and the relative inter-sensor scaling, is irrelevant due to the freedom that will be allowed later in the analysis. + +\begin{equation} +s_i = p_i \odot e +\label{eq:VK} +\end{equation} + +where $s$ is the simulated required sensor catch for achromacy (with the index $i$ denoting the sensor), $p$ is the sensor \gls{SSF}, and $e$ is the measured emission from each adapting surround. The CIE 2006 10$^{\circ}$ observer fundamentals were used, and the results are visualised in Figure \ref{fig:LSsimdata}. \begin{figure}[htbp] \includegraphics[max width=\textwidth]{figs/LargeSphere/LSsimdata.pdf} @@ -387,7 +406,7 @@ \section{Conclusion} Two methods of analysis for the \citet{macdonald_chromatic_2013} data are presented here. -The chromaticity-based analysis showed that there was a strong correspondence between the patterns of the chromaticities of the adapting fields and the pattern of responses. There were a small number of outliers, which could plausibly be due to additional inputs to the adaptive process, however these outliers were in line with the amount of noise in the data, and confounded by many other variables and sources of noise. This analysis provides no basis for rejecting the null hypothesis. +The chromaticity-based analysis showed that there was a strong correspondence between the patterns of the chromaticities of the adapting fields and the pattern of responses. There were a small number of outliers, which could plausibly be due to additional inputs to the adaptive process, however these outliers were in line with the amount of noise in the data, and confounded by many other variables and sources of noise. This analysis provides no basis for rejecting the null hypothesis that cones and rods are the sole responsible agents in adaptation for the studied retinal locations and conditions. The spectrum-based analysis showed that the results for one observer could not be well fitted by a simple model of Von-Kries-type observer, but that simple linear combinations of the responses of such an observer could be made to fit the recorded data very well. Though the ability of these fits is to be expected from this type of post-hoc fitting, the types of models which are predicted by the fitting align well with our understanding of post-receptoral signals, which suggests that we may be measuring adaptation at these levels. It would be particularly valuable to see whether the temporal nature of responses also aligned with our understanding of the timecourses of these different signals. It should be noted that this result could be due to the nature of the experimental set-up; observers were unable to modulate the cone activations directly. For example, to make the stimulus more green, the observer would inherently have to make it less red. It is plausible that this may account for the relationships seen in this analysis. @@ -420,6 +439,12 @@ \subsection{Further Work} \item For the spectrum-based analysis, currently only a single colorimetric observer is used (Stockman-Sharpe 10deg). It would be possible, and more correct to use specific observers relating to actual ages and visual fields. It may also be possible to use sharpened spectral sensitivities (see \citet{finlayson_spectral_1994}), though this would need to be done particularly carefully, considering the already large potential for overfitting. \end{itemize} +\section{Interim Summary} + +The experiment reported in this chapter aimed to extend our understanding of colour constancy and chromatic adaptation, specifically asking whether there was a melanopic influence. No clear effect for a melanopic influence was found, though the absence of an effect could not be authoritatively be confirmed. + +A large number of limitations were identified with this experimental set-up, and it was deemed appropriate to develop a second version of this experimental set-up, and perform a further experiment. This further experiment is reported in the following chapter. + diff --git a/LitReview.tex b/LitReview.tex index 4ca7ab5..6f9ec12 100644 --- a/LitReview.tex +++ b/LitReview.tex @@ -29,4 +29,14 @@ \section{Colour Science} \input{MathMethods.tex} +\section{Interim Summary} +This chapter has laid out the state of the art in the research areas which the other chapters of this thesis build upon. + +Chapter \ref{chap:Interviews} builds upon our museum lighting knowledge by filling the gap in our understanding of how museum lighting is actually thought about and selected currently, and tries to identify the most fruitful avenue for future research which will allow the reduction of damage to objects in museums. + +Chapters \ref{chap:LargeSphere} and \ref{chap:SmallSphere} target the interaction between colour constancy and \glspl{ipRGC}, using some of the methodologies covered in the Section \ref{sec:methodsforCC} and the knowledge of the physiology of \glspl{ipRGC} outlined in Section \ref{sec:ipRGCs} to build upon the experiments summarised in Section \ref{sec:ipRGCbeyond}. + +Chapter \ref{chap:Tablet} investigates a new methodology (a variant of achromatic setting, described in Section \ref{sec:methodsforCC}) for performing colour constancy experiments, using museum spaces as test spaces due to their well controlled lighting environments. + +Chapter \ref{chap:Melcomp} takes a more theoretical look at the relationship between colour constancy and \glspl{ipRGC} using a computational methodology, and the mathematical methods described in Section \ref{sec:math}. diff --git a/Main.tex b/Main.tex index 45f3bfc..e53d223 100644 --- a/Main.tex +++ b/Main.tex @@ -61,14 +61,14 @@ \include{Introduction} \include{LitReview} -%\include{Interviews} +\include{Interviews} \include{LargeSphere} -%\include{SmallSphere} -%\include{TabletMethod} -%\include{melcomp} +\include{SmallSphere} +\include{TabletMethod} +\include{melcomp} -%\include{Conclusions} -%\include{Appendices} +\include{Conclusions} +\include{Appendices} \bibliography{BIB} % This line manually adds the Bibliography to the table of contents. diff --git a/MathMethods.tex b/MathMethods.tex index 387c547..9369ffd 100644 --- a/MathMethods.tex +++ b/MathMethods.tex @@ -1,4 +1,5 @@ \section{Mathematical Methods} +\label{sec:math} A small number of mathematical methods which may not be familiar to the reader are used within this thesis. They are outlined below. @@ -46,4 +47,6 @@ \subsection{Principal Component Analysis} \includegraphics[max width=\textwidth]{figs/LitRev/Judd.pdf} \caption{The mean and first four characteristic vectors of \citet{judd_spectral_1964}.} \label{fig:Judd} -\end{figure} \ No newline at end of file +\end{figure} + +\clearpage \ No newline at end of file diff --git a/MethodsForCC.tex b/MethodsForCC.tex index 042120d..6dff3c0 100644 --- a/MethodsForCC.tex +++ b/MethodsForCC.tex @@ -1,4 +1,5 @@ \subsection{Experimental Methods for Colour Constancy Research} +\label{sec:methodsforCC} \begin{citequote}{hurlbert_colour_2007} \emph{How is colour constancy measured?} With difficulty. diff --git a/SmallSphere.tex b/SmallSphere.tex index 24c86af..87b0e9c 100644 --- a/SmallSphere.tex +++ b/SmallSphere.tex @@ -6,7 +6,11 @@ \chapter{Small Sphere Experiment} \section{Summary} -This experiment was performed to develop upon the Large Sphere experiment (Chapter \ref{chap:LargeSphere}) by narrowing down the number of variables and more directly exploring the question of whether melanopsin plays a role in colour constancy. Observers were adapted to one of two perceptually metameric conditions, one with a high melanopic content and one with a low melanopic content. Statistically significant but low magnitude differences were found between conditions for two out of three observers. For the two observers who performed repeated trials, inter-trial variability was high, and of a similar magnitude to inter-condition differences. One observer provided drastically difference responses during the repeat sessions; hardware issues which may cause such a difference are discounted, and remaining hypotheses for what may have caused such a large distinction are discussed. +This experiment was performed to develop upon the Large Sphere experiment (Chapter \ref{chap:LargeSphere}) by narrowing down the number of variables and more directly exploring the question of whether melanopsin plays a role in colour constancy. Observers were adapted to one of two perceptually metameric conditions, one with a high melanopic content and one with a low melanopic content. + +It was predicted that if \glspl{ipRGC} played a role in chromatic adaptation or colour constancy that we would different achromatic settings for the mel-low and mel-high conditions. No prediction was made regarding the magnitude or direction of the effect. + +Statistically significant but low magnitude differences were found between conditions for two out of three observers. For the two observers who performed repeated trials, inter-trial variability was high, and of a similar magnitude to inter-condition differences. One observer provided drastically different responses during the repeat sessions; hardware issues which may cause such a difference are discounted, and remaining hypotheses for what may have caused such a large distinction are discussed. This study was approved by the \gls{UCL} Ethics committee (Project ID Number: 9357/003), application attached as Appendix \ref{app:ethics3}. Code and data are provided: \url{https://github.com/da5nsy/Small-Sphere}. @@ -29,7 +33,7 @@ \section{Materials and Methods} \subsection{Hardware} -The sphere used in this experiment was 400mm in diameter, with ports of similar functions to those in the Large Sphere. On one side there was a padded port for an observer's face. Mirroring this was a circular aperture of 52mm diameter (giving a viewing angle of roughly 6.6 degrees), through which an LCD screen was visible. At the top of the sphere was a port through which adapting illumination was provided. An additional port, on the observer's side of the base, was added such that the illumination provided to the sphere could be unobtrusively monitored throughout experiments. A schematic is shown in Figure \ref{fig:diagram}, and a photo of an observer in position for the experiment is shown in Figure \ref{fig:SSphoto}. +The sphere used in this experiment was 400mm in diameter, with ports of similar functions to those in the Large Sphere. On one side there was a padded port for an observer's face. Mirroring this was a circular aperture of 52mm diameter (giving a viewing angle of 6.6 degrees), through which an LCD screen was visible. At the top of the sphere was a port through which adapting illumination was provided. An additional port, on the observer's side of the base, was added such that the illumination provided to the sphere could be unobtrusively monitored throughout experiments. A schematic is shown in Figure \ref{fig:diagram}, and a photo of an observer in position for the experiment is shown in Figure \ref{fig:SSphoto}. \begin{figure}[htbp] \includegraphics[max width=\textwidth,center]{figs/SmallSphere/diagram.pdf} @@ -62,7 +66,13 @@ \subsection{The Sphere} \subsection{The Screen} -The screen was offset by roughly 150mm through a short black paper tube in order to limit the interference (either way) between the screen emission and the sphere illumination. +The screen was the same screen as used in the Large Sphere experiment, but was offset by roughly 150mm through a short black paper tube in order to limit the interference (either way) between the screen emission and the sphere illumination. This was seen as a great improvement over the Large Sphere experiment, where no such effort was made. Analysis showed that at the average pixel values of the selections made by observers there was negligible screen chromaticity difference between the case where the internal sphere illumination was on and where it was off. However, at levels in line with the darkest matches made by observers, it was found that there was clear, but relatively minor, intrusion. The level of this intrusion is illustrated in Figure \ref{fig:SSA}\footnote{This intrusion was only found after data had already been collected and so no attempt was made to rectify this situation. In future experiments a more effective foil might be made by using a longer tube, either folded in the manner of a traditional light trap, and/or lined with a less reflective material such as velvet.}. + +\begin{figure}[htbp] +\includegraphics[max width=\textwidth,center]{figs/SmallSphere/ScreenShieldingAnalysis.pdf} +\caption{Chromaticity values measured from the screen during a number of characterisation procedures. All channels were set a pixel value of 25, in line with the lowest value that an observer picked as an achromatic match, as this is where the effect of intrusion of the light from the sphere to the screen was likely to be greatest. The black asterisk represents the dark condition (no internal sphere illumination) and each red point represents one of the primary characterisation measures (with internal sphere illumination as it would be during each experimental session).} +\label{fig:SSA} +\end{figure} \subsubsection{Characterisation} @@ -100,6 +110,14 @@ \subsection{The LED Rig} The arduino script set the \glspl{LED}, via pulse width modulation, to the output levels decided in the perceptual nulling segment of the experiment. The two modes had either the combination of UV and amber, or blue and red, allowing for the chromaticities falling upon the lines shown in Figure \ref{fig:LED_cross}. The combinations UV and blue, or amber and red, were never used. +An example of the \glspl{SPD} of the illumination inside the sphere under different conditions (following perceptual nulling, described in Section \ref{sec:null}) is shown in Figure \ref{fig:LED_SPDcontrast}, for both conditions for observer HC. These conditions were perceptually metameric but had melanopic contrast of 309\% (for this observer). + +\begin{figure}[htbp] +\includegraphics[max width=\textwidth,center]{figs/SmallSphere/LED_SPDcontrast.pdf} +\caption{The \glspl{SPD} of the two conditions for observer HC.} +\label{fig:LED_SPDcontrast} +\end{figure} + \begin{figure}[htbp] \includegraphics[max width=\textwidth,center]{figs/SmallSphere/LED_cross.pdf} \caption{The chromaticities of the Small Sphere \glspl{LED} in CIE 1931 space, with lines connecting the pairs which were activated simultaneously. From each pair, illumination with chromaticity values extending along each line was generatable. Even if there were differences in the spectral sensitivities of the observers (compared to the CIE 1931 observer), with these narrow-band primaries there should theoretically always be a point at which the two lines cross.} @@ -131,7 +149,7 @@ \subsubsection{Perceptual Nulling of Peripheral Adapting Field} \label{sec:null} The basic logic of this experiment is as follows: under the null hypothesis two adapting fields of identical appearance should cause an observer to be adapted in exactly the same way. In order to design two adapting field illuminants which appear identical (perceptual metamers) we can either make predictions based upon standard observers (with parameters set to match our real observers regards age and pupil dilation etc.) or we can employ a process whereby individual observers make minor alterations to two conditions (that are designed such that they could be metamers) until they appear identical. We have opted to use colorimetry as a starting point to choose LED primaries (Figure \ref{fig:LED_cross}), and then allow observers to fine-tune this matching. -We run the risk of falling into circularity here: we ask observers to set two fields such that they appear identical, and then (in a roundabout way) we ask them whether there is any visual difference between them. If melanopsin does have a direct impact upon visual perception we are at risk of accounting for this at this stage. In an attempt to avoid this problem, we perform the perceptual nulling under conditions which we predict should not allow for \gls{ipRGC} involvement, or should minimise such. It seems to be the case that \glspl{ipRGC} do not react strongly over very short timescales ( faster than roughly 0.5hz \citep{spitschan_human_2017-1}), with cones being much more active in this temporal window, so we chose to alternate rapidly between the two conditions and ask observers to make alterations until there is a minimal visible flicker. +We run the risk of falling into circularity here: we ask observers to set two fields such that they appear identical, and then (in a roundabout way) we ask them whether there is any visual difference between them. If melanopsin does have a direct impact upon visual perception we are at risk of accounting for this at this stage. In an attempt to avoid this problem, we perform the perceptual nulling under conditions which we predict should not allow for \gls{ipRGC} involvement, or should minimise such. It seems to be the case that \glspl{ipRGC} do not react strongly over very short timescales, as discussed in Section \ref{sec:ipRGCs}, with cones being much more active in this temporal window, so we chose to alternate rapidly between the two conditions and ask observers to make alterations until there is a minimal visible flicker. Observers were instructed to fixate upon a small fixation point displayed at the centre of the otherwise dark display. They placed their hands upon three dials which could be independently varied to change the peripheral adapting illumination inside the sphere. @@ -159,7 +177,9 @@ \subsubsection{Achromatic Selections} 3 observers took part in the main experiment (the author (DG), HC, and LW). -The task was identical to that performed in the Large Sphere experiment; using two sliders (controlling the yellow/blue component and the red/green component) to set the foveal stimulus to appear achromatic. The instruction was given to set the central disc such that it appeared not red, green, blue or yellow. As previously noted, instead of 10 runs of L* 85 descending in 5* increments to L* 10, there were 30 runs of a pseudo-randomly permuted (each run) set of stimuli specified such that there was one at every 10 L* interval between 30 L* and 70 L*. The range was reduced so as to minimise the impact of gamut-boundary issues. The L* interval was increased to allow for a greater number of repetitions of specific values within a similar time frame. The order was pseudo-randomly permuted to avoid any trend based effects. +The task was identical to that performed in the Large Sphere experiment; using two sliders (controlling the yellow/blue component and the red/green component) to set the foveal stimulus to appear achromatic. The instruction was given to set the central disc such that it appeared not red, green, blue or yellow. Once happy with their selection the observer was to press a button, at which point they would be presented with a new random colour. The same starting condition as was used for the large sphere was used. + +As previously noted, instead of 10 runs of L* 85 descending in 5* increments to L* 10, there were 30 runs of a pseudo-randomly permuted (each run) set of stimuli specified such that there was one at every 10 L* interval between 30 L* and 70 L*. The range was reduced so as to minimise the impact of gamut-boundary issues. The L* interval was increased to allow for a greater number of repetitions of specific values within a similar time frame. The order was pseudo-randomly permuted to avoid any trend based effects. Two observers (HC and LW) performed 4 complete runs each, 2 under each condition. The author completed 2 runs, one under each condition. Each observer only completed one run per day. There was a gap of roughly three months between the initial runs of HC and LW and the repeat runs. @@ -220,61 +240,61 @@ \subsection{Primary Data} \begin{figure}[htbp] \includegraphics[max width=1.2\textwidth,center]{figs/SmallSphere/LW.pdf} -\caption{As per Figure \ref{fig:SS_DG} but for the data of observer LW.} +\caption{As per Figure \ref{fig:SS_DG} but for the data of observer LW. Note the interaction with the gamut boundary, most easily seen in the top right of the middle subplot, where the lines in L* break down.} \label{fig:SS_LW} \end{figure} All combinations of sets of data were submitted to a two-dimensional two-sample Kolmogorov-Smirnov test\footnote{Using the function available from \url{https://uk.mathworks.com/matlabcentral/fileexchange/38617-kstest_2s_2d-x1-x2-alpha}.}. There was a statistical difference for all combinations ($\alpha$ = 0.05) except LW-AU-7-20- v. LW-RB-7-21-, and LW-AU-10-11 v. LW-RB-10-12. With Bonferroni correction to account for multiple tests ($\alpha$ = 0.05/45 = 0.0011), HC-AU-10-18 v. HC-AU-7-21- additionally fell above the $\alpha$ threshold. +\clearpage + \subsection{Secondary Data} -Data describing the characterisations of hardware follow. Figures \ref{fig:SSLEDs} and \ref{fig:SSLEDs2} show the chromaticities of the adapting surrounds as recorded during each session. -% +Data describing the characterisations of hardware follow. Figures \ref{fig:SSLEDs} and \ref{fig:SSLEDs2} show the chromaticities of the adapting surrounds as recorded during each session. Figure \ref{fig:SSgamut} shows the recorded gamut and white points of the display measured before or after each session. Figure \ref{fig:SScal2} shows representative results from one of the secondary characterisation sessions. + \begin{figure}[htbp] \includegraphics[max width=0.9\textwidth,center]{figs/SmallSphere/OOwide.pdf} \caption{The \gls{CIE} 1931 chromaticities of the adapting field as measured during data collection, with the spectral locus for context. See figure \ref{fig:SSLEDs2} for further detail.} \label{fig:SSLEDs} \end{figure} -% + \begin{figure}[htbp] \includegraphics[max width=0.9\textwidth,center]{figs/SmallSphere/OO.pdf} -\caption{As per Figure \ref{fig:SSLEDs} but with different scaling and labels.} +\caption{As per Figure \ref{fig:SSLEDs} but with different scaling and additional labels.} \label{fig:SSLEDs2} \end{figure} -% -Figure \ref{fig:SSgamut} shows the recorded gamut and white points of the display measured before or after each session. -% + \begin{figure}[htbp] \includegraphics[max width=0.9\textwidth,center]{figs/SmallSphere/SSgamut.pdf} -\caption{The display gamut at multiple pixel output levels, and white points, for all sessions.} +\caption{The display gamut at multiple pixel output levels, and white points, for all sessions. Measurements are grouped by pixel value (0-255).} \label{fig:SSgamut} \end{figure} -% -Figure \ref{fig:SScal2} shows representative results from one of the secondary characterisation sessions. -% + \begin{figure}[htbp] -\includegraphics[max width=0.9\textwidth,center]{figs/SmallSphere/SSseccal.pdf} -\caption{Representative results from a single secondary characterisation session.} +\includegraphics[max width=0.9\textwidth,center]{figs/SmallSphere/SSseccal2.pdf} +\caption{Results from two secondary characterisation sessions. There appears to be a minor rotation in chromaticity space. Importantly, the distortion appears systematic across sessions, and therefore is unlikely to be the cause of the large difference between the two sets of LW data. The level of differences shown here mirrors that measured in all other characterisation sessions.} \label{fig:SScal2} \end{figure} -%\clearpage +\clearpage \section{Discussion} There were statistically significant differences between the achromatic settings across almost all conditions (exclusions previously noted), including nominal repeat conditions. Most of these differences were relatively minor in magnitude, and within the range of what might reasonably be expected from various inherent sources of error and/or noise. -For one observer (LW), there were not significant differences between the different conditions, but there was a large magnitude difference between the repeat conditions (See Figure \ref{fig:SSsummary} or \ref{fig:SS_LW}). To be clear - with a single day separating inter-condition trials, this observer exhibited no difference between conditions, but after roughly three months, when returning to do a second pair of trials (again with a single day separating) the results were strikingly different. This is particularly unusual considering how well matched the inter-condition responses are; if it were just the case that this particular observer was particularly unreliable in their selections (and their data does exhibit generally quite high standard deviations, see Table \ref{tab:SS}) then I posit that they would be as unlikely to be able to repeat their selections after a break of a day as they would be after a break of three months. +For observers DG and HC there was a consistent shift to higher values of a* under the RB (mel-high) condition, which corresponds to an NCS value of R20B (red with a small amount of blue) according to \citet{derefeldt_transformation_1986}. In the case of observer DG this was accompanied by an upward shift in b*, which suggests a shift to a yellower red. + +For observer LW there were not significant differences between the different conditions, but there was a large magnitude difference between the repeat conditions (See Figure \ref{fig:SSsummary} or \ref{fig:SS_LW}). To be clear - with a single day separating inter-condition trials, this observer exhibited no difference between conditions, but after roughly three months, when returning to do a second pair of trials (again with a single day separating) the results were strikingly different. This is particularly unusual considering how well matched the inter-condition responses are; if it were just the case that this particular observer was particularly unreliable in their selections (and their data does exhibit generally quite high standard deviations, see Table \ref{tab:SS}) then I posit that they would be as unlikely to be able to repeat their selections after a break of a day as they would be after a break of three months. The differences cannot readily be explained by hardware issues; variation in the surround illumination and screen output are minimal between sessions. Further, the same level of hardware variations would have been present for both observer LW and observer HC, and no such dramatic shift is seen for HC's data. Regarding Figure \ref{fig:SSLEDs2}, which shows the chromaticities of the peripheral illumination, it can be seen that there was some variation between repeated conditions, and also a small amount of variation within each session. The variation across sessions is likely due to physical disturbance of the equipment (either the measurement device or the illumination source) and is of a relatively minor magnitude. The variation within session seems to be systematic, perhaps relating to warm-up time of the illumination source, with the chromaticity being quite variable for the first few minutes of several trials before gaining stability. -Figures \ref{fig:SSgamut} and \ref{fig:SScal2} show the results of the primary and secondary characterisation protocols respectively. Figure \ref{fig:SSgamut} summarises all the collected data, showing that at very low pixel values there is a moderate level of variability, which seems to correspond to the sphere illumination, suggesting that some illumination from the sphere does reach the screen. Figure \ref{fig:SSgamut} also shows the recorded white point of the screen for each characterisation, but these points stack fairly precisely atop one another. Figure \ref{fig:SScal2} shows a representative set of results from the secondary characterisation protocol (for observer `LW', under red-blue adapting illumination, initial trial - 7/21). It can be seen that there is some distortion between the values as recorded (following calibration) and the measured screen output. There are several potential sources for this error; the calibration routine (for the calibration we rely on tristimulus values computed directly by the \gls{PR650}, whereas for the measurements we record a spectrum and then compute our own tristimulus values), the experimental code (there are a large number of colour-space conversions in the experimental code, it is possible that one has a minor error), amongst others. Importantly, the distortion seems to be systematic (a slight rotation of points roughly around the white point) and of regular magnitude across sessions. +Figures \ref{fig:SSgamut} and \ref{fig:SScal2} show the results of the primary and secondary characterisation protocols respectively. Figure \ref{fig:SSgamut} summarises all the collected data, showing that at very low pixel values there is a moderate level of variability, which seems to correspond to the sphere illumination, suggesting that some illumination from the sphere does reach the screen. Figure \ref{fig:SSgamut} also shows the recorded white point of the screen for each characterisation, but these points stack fairly precisely atop one another. Figure \ref{fig:SScal2} shows two representative sets of results from the secondary characterisation protocol (for observer `LW', under red-blue adapting illumination, both trials). It can be seen that there is some distortion between the values as recorded (following calibration) and the measured screen output. There are several potential sources for this error; the calibration routine (for the calibration we rely on tristimulus values computed directly by the \gls{PR650}, whereas for the measurements we record a spectrum and then compute our own tristimulus values), the experimental code (there are a large number of colour-space conversions in the experimental code, it is possible that one has a minor error), amongst others. Importantly, the distortion seems to be systematic (a slight rotation of points roughly around the white point) and of regular magnitude across sessions. This suggests that there must be some other cause for the difference between LW's first set of data and the repeats. Since it is not the focus of this investigation I shall not dwell upon this point, but I shall suggest two possible causes which have presented themselves. -The first - although relatively clear instructions were given to observers, it is possible that observer LW chose a distinct interpretation of instructions for the second set of measurements. Specifically, it is possible that the observer switched between making a `hue / saturation / brightness match' to making a `surface-colour' match. It does seem to be the case that for the second pair of datasets, the chosen white points are considerably closer to the chromaticity that a neutral surface would possess, so much so that some responses from this observer hit the display gamut (this is what causes the degradation of data visible in the top right hand corner of the middle plot of Figure \ref{fig:SS_LW}). +The first - although relatively clear instructions were given to observers, it is possible that observer LW chose a distinct interpretation of instructions for the second set of measurements. Specifically, it is possible that the observer switched between making a `hue / saturation / brightness match' to making a `surface-colour' match. It does seem to be the case that for the second pair of datasets, the chosen white points are considerably closer to the chromaticity that a neutral surface would possess, so much so that some responses from this observer hit the display gamut (this is what causes the degradation of data visible in the top right hand corner of the middle plot of Figure \ref{fig:SS_LW}). If this were the case, it would underscore the importance of clear and intentional observer instructions (See \citet[p.679]{foster_color_2011} and \citet{arend_simultaneous_1986} for discussions of instructional effects). The second, though I am aware that this sounds rather fanciful, is that this result could represent a genuine shift in the perception of the observer between sessions. Though it seems unlikely that such a drastic change could result from such, in the absence of other options I feel it is worth considering. There is evidence to suggest that an observer's white point may change seasonally \cite{welbourne_human_2015}, as an effect of changing chromatic distributions of one's surroundings. Three months between sessions seems like a reasonable time-frame within which to witness such a change, particularly when the difference is between July and October (Summer to Autumn), where it is most likely that there would have been a considerable difference in the lushness of vegetation. @@ -284,6 +304,12 @@ \section{Discussion} The data of observer DG (the author's data) exhibits a lower level of variance (probably due to increased level of familiarity with the test) and exhibits an inter-condition difference of similar magnitude and direction to that for observer HC. +It would be desirable to understand the effect of luminance or L* on the achromatic settings made, but as with the analysis performed for the Large Sphere data, there would be no means by which it would be possible to separate out the effect of the experimental set-up (whereby the sliders represented higher chromaticity shifts per unit movement at higher values of L*) from any underlying effect. For this reason such an analysis has not been performed. + +Colour constancy indices have not been calculated for this data for a number of reasons. Firstly, there is no meaningful white point which could be used as a `pre-adaptation' point in such calculations. Secondly, the `ideal match' values (which would under normal circumstances be the peripheral adapting illuminant chromaticities) are designed in such a manner that they are metameric but also have a large difference in chromaticity space (Figure \ref{fig:SSLEDs2}) due to the difference in \gls{SSF} in the fovea and the periphery. For this reason, a chromaticity based description of the surrounds is likely to be both misleading and disingenuous, given our lack of knowledge about the \glspl{SSF} of peripheral vision. + +It is for this reason also that I have been careful not to plot the chromaticities of the surrounds alongside the chromaticities of the central achromatic settings. + \subsection{Limitations} One key limitation in this study was that there was no prior prediction of the direction, nor magnitude, of the difference that we expected to see as the effect of changing the melanopic activation. This stems from the fact that though several researchers have sought to examine whether there is a link between melanopsin and colour constancy, none have proposed a clear theory for how or why such involvement might exist. @@ -293,3 +319,11 @@ \subsection{Limitations} If there was a model which predicted a specific difference with a predicted orientation and magnitude, it may be possible to use alternative analysis methods which would provide greater clarity. Further, it may be possible to design experiments to more robustly test a more specific hypothesis. Chapter \ref{chap:Melcomp} aims to fill this gap in our understanding. The other key limitation, which applies equally to this experiment and the Large Sphere Experiment, is that there is an implicit assumption that foveal adaptation is affected by peripheral stimulation. Though this is implicit in all \glspl{CAT} (with only a single transformation applied across an image), it is not clear whether there is a physiological mechanism by which a single transformation could be applied across the entire retina. Indeed we know, from simple experience with after-images, that adaptation to a small area of the retina is certainly possible. It remains unclear what the level of adaptational cross-talk across the retina is. To minimise the impact of this issue, and to increase our understanding of this process, it is recommended that further experiments of this type employ a modified design such that both the periphery and fovea (and further divisions) can be used as both adapting field and test field. Further proposed modifications will be discussed in the final chapter of this thesis. + +\section{Interim Summary} + +As with Chapter \ref{chap:LargeSphere}, the goal of this experiment was to explore whether there was a melanopic influence to colour constancy or chromatic adaptation. As with Chapter \ref{chap:LargeSphere} no clear effect was found. + +However, it was noted that there were a great deal of assumptions about how a melanopic input might operate, which were implicit in the experimental design, which were not necessarily backed up by what we know about \gls{ipRGC} physiology, or ecological requirements. + +It was therefore decided that further work was required to understand the ecological requirements for colour constancy, and to work out abstractly whether a melanopic input might be of value to colour constancy. This inspired the work presented in Chapter \ref{chap:Melcomp}. Alongside this, an experimental method was explored which would allow colour constancy experiments to be performed in a wider range of environments, including more ecologically valid ones, than had previously been used (Chapter \ref{chap:Tablet}). diff --git a/figs/SmallSphere/LED_SPDcontrast.pdf b/figs/SmallSphere/LED_SPDcontrast.pdf new file mode 100644 index 0000000000000000000000000000000000000000..a90783a3a79a8a0d3a340b5ccf850f66ec9777e8 GIT binary patch literal 20135 zcmX_GWmp_dv&A8}yDaYR?(P!Ybr*Mc4ess|2o{1{upo;EcXuba!^eAn+<*OaojO(3 zb$VucW~kMqWSH4mxDctUs%vTy*#T?-Cv#gwAt6LoHCHD~4+|?-0G+tAnFY`aAfu`V z(6(}Q135VYI9Rw?*qAj{#MLzARJ8;EpEOcV79I{(j_$IqX3jv6g&T(yJtC{5tCgAi zr-PK4yA^;=N`Qlnou8e9ot>S7n~R-KpB_;}1kuXT^1qlI{~sH`s$}J8;|>IHeokos zSY<%=?w@(E%GiIhNLg7pSy}-;{oLF?$sG{A^P2NV5}5Q{3m)#U>#6Yag>HUfbB6{) zBr}B|#h}Af$}4=M0EibEn|#<HE595~yI(?*4N8qWXOG@d{cZ17AVE-;AjWUu{;rwNk&|zO#P3zZAZoR0K`wysHF# zwB9%F@_sBjkp;cX76x@cpgpZmlQrw0R2*-<9v=sp!%X>8iTX#-EVl$)Tn0X$_DuhM zUsc_|vhlt7KnMNnd3qET-M@W&5Pcpsc@OG#t*9GBTZq#s44S&vne@9X7_0kNx%Q${ z*W5Cf;A0qcdm8u?`>kQ41uP=s^)_Ai+CJ~d=<_sE_rh)r?npP@vB)fZ?I5#j4@ghb zV4)^^^?I5J#1BajPIT$YZnfD z-@jT6DhqoAC-(TfzS@YsofbN%Jm6iP)Y)2PY@QCjdHv>FOo?~XCVhGEar-cmoa8(g z1+qV161vIz1wOUxJ-z()dHu9+6%gbu(6i#m zTnZS~nX!ABgD*(ZzVMlRtF3P7E^Ase6vfY#;{-W_mX})MSSoGd05jv&P60=P;FccS zRENfecl_0bU#m)X{Eg}5i-|El0{l{xFAeE+-TFT*&??fn(N%7UWNYh`;vJgWFG>p< z*sI?~s{Cj<>J*Hygef&<)eHz;1i{)p*W-k1j${_8Kb9&F*!ii??XcOU8zGglK?0n1 z2lk2``jwaxv+@G+rB5RMtG`|r7Xn+-pTWPri!|TVK7cH8Ev)oX;_FfPl?2B;hleQ5 zm)hMJR8o91@vge^im;fnGgscyfJ>*0R&PqXEoO`A9rTH`R3jByuFnh9>B1#=iubRk zz&30*98t06OKsT(ocR#quNswshxI1W6UjPkBAy0i%$1t4weh6bth^#UR8UkkR7u@# zl6N>uin-jPG8sKBFcW8u0%Z5U6$ucozjo_0SH}-%rieF=Q)Qo)WtST_c58mrF!1O2 z8ddrucLs95f3tBys+!1<`=f~gs^Hlxd(zL$e;D>W!Gkgbksb?1q<3>f{g!OUC+p8> zZx)CuBk>2!cC6TXz3s4222@MU=#UbZJE(TcXqZI^T9jSBaH$>_0@G14zNBY%I(ntn`4I@-r?QPLha3jZcY z!jg#G=N0b&blDq~ruyJ$p{LXy^@-4lz+BV!fC^n{C7{#I9=#{#TZ) zhGq#<1i}PW;5j&bc_&;ex&#qU$l<8dx=i(_F`HN9g;9}aaKG2eWLpDftK}>v(9GsXAgOj6rdB*$i7f!-<{??f*w(m*M{EB0Cj5fOZ8W4hO zjB>7Zqpbqp`7N%!mOVvv`KNWKDRSpgsu(mow8`?cKM)(>ksm`f`8r@!d4X+-Ax5a3M0*B&xQr92*Hp>_- z(9l{ZK-uiNl{1N3Djxdj6`DJULX8t@j z2gQ^T>@xCI!T1yPY{P~*+20y5YAyPdEw}ktG&cFITKyH?IJJLyDhA`qR6Mh#LI{k+ zv8=zKN#*)CrIg3jc})4b#J&>28xJhD)7BnT6tL8|Qc&3{iBOsB9j>I{k(`F8PrEv2 zp(eZiATBH7+RfW`b0xh;&SN3!2xLes3S$B2`?O5sp&8hIRKxMxe{q-`1Z-UMr# z71WC0l}Q`1eLptzChCOuaVMVW6Jzpx`E%$8OHNg z-8ldAb5u6=n6jRBtL929Z|iqpG^DO{4}~8ph-R*-+@hZzYFWqM?qTX0o0O*(`U$pn zt)G8fb7p7GA;oZLV~>uaCpl@7*-)WTbs2jA^_tsmu($HXokm5~oV?8;xk;XB&s#GKWN(We%Sdmc0Q}2XjdEdm; zywHzti|w;F&AW|+XjTSD9QFaAaSAgrp0|P;r8X@6U4VVuwr%1avp%CZ=Ha)dS36DG zy0e2skVHy0CTI6PJuknZh{AMBdBwxHWO)oxfT!~6?*XZ0%uEi34wXO;gLtJ@T@%mp zx)_+kz`uBnUr~cb%jt6xa;zEPh{X*&Vk439l&@6{wyI}-1?&wEpK#=~h}~ztaKt0u z07w@6_4dHWFxkWU zr_>>guxqjN>6Vx+Mj5c;m~hmPesVa;k++3eaXO>VC;uv)j8t11D*u?4G|zqa(@%mc zp&*ZwA|&)<3>-B>T}GaWvtVw2JVhXqI1T3!xS(3~#*`^@e9;X|lTIz0VsuSc=7=-Fd0U0^ybA^UPFaBLM6vMFg-{(Q`39N=&Xmtjc@DW#axy8MFwppyX( z`JPn+vx~ka9LtN2QWcQ&YoyTAy8K&G$t^p*`w}{x;^}LY@>>{KO>U^piCJW>kzbV6 z;XR$5T(6Z?s_CiO=#+ zO)tsJWY}dhbYv~jR|mIx$tUAcdz!Rb@Ad# zl&<2Wf`3*(=?c4V}8{!Z_#}^_wPLZDPZ_wW|IatXi2vlPVHg zZV$@kIY;U;q>FZ3sU|k4 zP?bAbX-JD0pVp?%43Mml(a;f*2vGYt?Nh~@E|Gtu%FiRuS#r9_Tq>L)m3)SMr6{0SF`X*)*2^aiUe|l1CX5z}o;!2HQx2OW*!O!@G%@$QDx79At1vrp z)3U>Ua0v}3XJMITT*f3>UfHRSLCg%C`(c%A=c|R()|L+gZ^#u(gPVccLFcL6}L=R9_wE(^8=!8a=DiD|f0Mt`+Rur*AnSVk)V&Y10$Nu{{W4)TCY z*Hj=r4!H zr$h-zU{|ZJCkK3$n*BiyVMU%cIitO-;F!m%C#Fdw5P>;mITxF6i<*O-a!Wp6wBl}S z>|7PEZDydu@-L<=D{C{;O`Wf!k}))K_(;*;y-3;-#KOa*A$KQLxQJPltoCn zK>hTT27*n$0JnB#AqH?-8IT}L2cfU(`X>!x3nNDjmiX{A z!>Degl_Ct6v+ciIpY9lO5gb4h>2bM;RZ{JoyS-LL7T3-n4yS(2uick0P36-SSnJAV z`YAvyr5Z_Rxqy=79doQlTf>^}3?csD(RF0^ z+SX5a<<9-w609pL7UOUK;;EIN7<0Hli_I+#&-K{Ki6dY7l_qLl#84mIq_XJL`w=N# z1zlnd&mF2!27D!<$28I$R$FA<<=oY8P5E6Y8t1;Kh)R~A=qN6_vnY=%+UA~&IX9jX&}Uy27>Q(snr`5ZA@Gh5 z)waO_uJDe~B2R}B{8{FK);3^3O~Rw`+eB7~lUR=D2Bum_P*XPzakMp3s*8K5VUm=( zXu^M&|kPOz(^UO!EIDYH4;$)i7|B5LF zjeRHXoZV8c`iMJ-YqIAjN`@hjyEqq1t)rZ*ng`D9kL)z=rkj;vSXs%;ZtAm&mlj|g z_7M-nkgS+j4$5hf=t&`hHCpsN3MLPh72%NmWl31=_biF3HC26WIxL(R!#@^6dGz-Akk_!z@@AFlkg_Ty@=4)Uf6E@m7Iz<{uo}uqI83=`xg~Hj8$|$-FzBJuw zTIQePZ3f8BHQD%IJm#WobO-QPj`y1)Ql@q)4lwB5mY$JwWSC&(ed7J`}qUzn)~)5{&a zCB;+fQzb%m2+45wzw`rlhT}sBxbgw<{64By)EDc_10~YfvTCy_vRw4TbLT7@jQe)v zsGp}SC&un5MLNIa3pAi>bG@GH6oS01Q|X$&qBeExEDqM{js-S$;=F5FJ78pYzSuz| zP5Rj5z8_Fo`w?3iB`fT)(?1x@Mcq$@840gQ?OGnif(VuGp;9Xi_!{IML5}HIkT?CU ze$<~@%8{$h|Lf+rW|GsGZNVV>sl9<+~%P0k-Ap^mvAm-Q7 zO*nOEiiUjY{yaL=7G7jZf1Gi|fGw*{4>FRCU_!-x+Q#N;f)~RKbeww%& zkW#LIO;N-Jlt9TUjb4O_m9r~{EnnXBNKDIY5RoFSwY3eQ zMD9)l#o}qN5qSODACfZ7MZRX4%_?)tR`!BNtV+a^u;Up?dhbhgd9#7v= z#ER8BAwoA6rZs~#VR-0j{K5!*l&24`Q&UgCiJf9WqRRREk_+fA0N=1A-nmxT=*@a1 zIoeZcr&NXVhBL1Z4o5L9UUf4}=``l$V!){(ygFPHJK^s^wqS+4lr0oDW~YW0_gZX- zi=O8lRiUxOFOK${vSM~cnLT0xeLdU8P6lEs>4A0b=lP|t3jA>k8zY8(j%pG|Kfy&1 zAwcz;??ZgkEl+L@J<$=;B#x9p+w#zIW5S=EQRMA3t-w={j36$`5`bq#M#{mk60R}M z=TR#s7mP&yC+;pHJrD5(Hqk=7UwWQD+oeNpv+b0f$w!vbL@A(&(nh}~o>fEke$w2; zD+f6q7gCWZRwXI(pV-U}D1h>gMZMKoxp=~zqt$+J80Faw3(5cZS2}R|7?Z8v>Q?H_ zEL^4!W}a(FNL*5WfrS1gA@20TLpj%B)S9o4nQPL)Bx58_pZ9s$>gu`OQwtQ@ z>)-#Wm(vMT*66z>z~t;3yETY>t(X8+o-DbAR?S;oNt99QT-Q&fDh(nN&Vp*|k`kPg z$(E%jhA9s8E#NShiGwIx>qZ;aRs%t-oL0H3Vq;ClQn0l|vL=^-_eV}C zV4;M@qS8#}i}R+053Xy0M_jTJI0wb+PMdMJ>r^yabChbV(uBL5hqvvqN)+$_CI_-{QvId|-c-IFGtJnQ6xv!<~VI)Y)#@xU3-L$b>>ni;nO739dd)nCGJL-I~_L+n@@ar*exS33?>`5()2oPeW_&Eo?_As69 zYP#{6BKz0J8kTUHA*0#q^Ei|}rN<*6TpFjbIS#c5Kxpm#oeV0ml$-EY%MtjJp5*{ne+buI7=Qz(;K^0zh;utrru zIVr0Xt<06KG}9h)>}B(}Bd)3|>+qb~ zTTaWC*u5P_6dP@dzFn7G@G#VS$S5N3lM8rKle#h|zUG;c<%qY9lX4BYwV1js?GJwa z89**fMPPe=@emLqBWZRKMfe@5J{#5y2YE$hIs7IGJ>^_`no3+@SOZl8fhr$iq{=XK z!HcUh9B8Th#xMWif*G&9^Yv6bl`nl2doWW#N%o4G%ZIy@1YTK%Z;+^aAWWjLwRwS| zE6oSi!6x(mT}7Tl@m*4~gKS=qWkd-5oQt`C1{Ax?b=+krB0At#GI0G%!LoeMETDpBrZ;4;yR(@EiA9=AtEJK1gcfu#F`BG_VyE7Y9d(ynC_O#0 zp&()NJ}S)pwaYixWxCi)PGG|h0(NAqI8N&FSQ(2)#<~BKMaE4%1ZLAdoOM%SeD^vH zO)Wi|rS8h|+d&}ALUFX&NLo~`$Wx&UG=XZDDxAGAaz@H;IkcRO!b8*{sl--$Tu}nm z-{YA+?$+?rz5@}cCe87HWKWROmtJuWkDaF2KBwft1-B`>a68%5ulx!zMR%;#UBYGR zlW7EpZ_8`Q} zg->gEVy{cc+>+rhncB6O!x@_u<8OeXAB2R9WNPh232$l4J=lc%N7H4P!n+XXucvQHlfu%IWWFS1J-ZX0l zx6&i0txGEFTb7yV846~fIjV`5pNcCb!+LRk9u?H=(&Sa%T@dk5K?dY&F;2AvgCi-4 zNgClW7rJMboI?O$TkYV}1 zP#$dcbr4-yKdx_Oq$2ZWDx|S_y#xCxTB(6QB};Sie;vl94=vXdYxf2$P{zT_Yo1QT z1da5KTQg%#<$xsr;$3I`9uoHfO%30)f1_Mcn(y{&EBmG{>lZf+kMbggqECxrZX3_R z)6O(-7gsK+Pk@dw1W96(3#nCl+=Db?pL}2i2r%qBtO%Uz(V2rOn5>;6i zKR2RePPJ4_YTMvHBVhk%tA4b5nUb*bd@Y|V2CH2`+6l}0>`{h)x@iH|*OklS9}@d| z91?Rl)Yoi&b%B`3EK&`&QU_i#kU7(3sLbDtgl3tx*^aIea+dt?%d--na1?%FTc+}{ z9&x92F_|bD@F@q)N9xrJqNJr1ZLC@f0|drqJEVkDSt5=E2S$o8+zfuE*jf;jSUkHe zpD5RGr7{;E{`A1xUXT|0LkH3woETH@F?~6?53)s9v@?eUB0$J^^gQr^utOl(%|I5y0Rgwu#EvnH=<$?@8kYSUGKQ zBZyUt%!Mzy*nPN$oq89YcIqEL`L@;d!l^!y8EwFO(tsIPBej8aVvq%XqK#w_WBo<@M5>2gSz2I3S%w!47U%dwh(6|bwE}l zBW$6j0By5QgU@GAlze)Ewhmxm_IJly`GM`Z+!BZ)&oIB^49#=AQ|J{Rl!jh zr50wWu~;Z+KckOr+$2iM!Lrd4uaZ%|oVa;4y}{21j=s1U2woPUj(`oBX_*j7Uq|IC7*`M3d*1!hn z(}w3Rd#hnf-?w(@&Cok5-LBFzt`!ICDS>U;DLa=ZcVgb-=^_UwBYs;}69q$gtVy1ve?y;w0~MHncxP|e&GlZ8w&~(;i-(=Gk10_KQQl0tRALOTAvl>0uV=b zeup|-mXcFD6ox^m)Fu{!bzSm$`i=cp`o0x}<`w-MWz@%aZ$ATb+f~spFhw zWit7^(LR`|LQ-o&C0P#}?j~gd=9NlG@P6a3*#tt0EavRH@F=0nGdCsU-K{-pkB>t+hq>r^*ENhQWcmIX^SC9bqbZ%G1 z2A9q?v!XqeDmLaKFO3?OgMiLNhjJ2L96M&Vbw(!$mp_2mDJSUH4&ZV~7v8tBYp8oI zV&zl%t0oA<3FdgK^&A4t)Ez{IP%AZYlO3Ff_G`5CV#1cr(Kv$D!WR;S(9+ve&7=2E zVI|h4w?OIR$v0etOf&(2=C>lY*KjLX2iiP9HrCu25vt3;>kr3Fgn2z?v{R8Ed^#)- zA=;Gk;(Zv(m|;n&6(o%5ufb}?dq94WdJR4oBSi{#@;SCj$6Ns7Fng;zXp=szcQOmo zJPxZebE-h=TUf$QqPnS6q{gaJ#az*(_d*zrBw>(3snop+;hm5Y?MlXb8S;mjvVrEz zqF^|cc+yLRhM?WJzHBWnHeM{|kEEs@dIc{kVU$%TRKGa$dj~Nr@!VStxkYG6rw~%T zVIIL>L&5YZU+;?M%XQ)U9okU*djHr7|0a;OF4BoV3YXU#X#xC_SA-*f)d*nc1x+Z7 zRSZRkW8qC4FF*&SN1Jk=XbrH=fa<@9`F*FmEfpdRA0(YVa%g@rwvsb*q#!@m^Q?tm z2*1-5`hicKM+>xD#aE!SASP| z{)txkWs<&DV_i?5OXR>eQk!!FeH%((?}sS4FNlO3ld84@)cKGJN|b?*`Cd=8mBD#e zj+~THypvCI2^X9pHktlxO0|cxhup@A7*Q2&N&Q=ES+igl<8zM=J0WouHx5Rn1=RuM z)tMX9K`e7bC$~|E6X(QQ@_IyE5e=2P<|7PYf6-G^O~DNO&dHG@Zq|49|r#bcL* zz@%F!PP96nW(HSo?K9>MYRrxRBp7 z)fVE0x9E6m*_PM*q3%)Wy+Rr}Uyt)Cx1e_zq315* z+v=7YeTPH5s8*Dwc&c_11V(Awyxf_WJHzT7gJpp*PpB(78Nbd{(|7|z>g=A)m?jVR z;`kiu79ZpFf>yH^Gq7QhxPP!TN)1=Z%2e%xf>I_pi|I>+P=7 zsiW{;{ynfNmf!bx-O8fMv{3$pc*p*(G@21FWh8n!ikPRSD4D{Yw=+F*M7t!$z^#}z zUnynKgR_rjXv#ooGg{EUqQge%n3x@(T1#gnz2g8CFo{ha;bsA_nG$JG^ixuNP$wV* zglu(T-g@1+h5C6VnN+A{%IrqjYjC8*n8uRzv$PVPm!w90=lD|`OP#47T4-g zvpH}Y2&TLx=E^A(1}UwY!G`_8^TC9~5v`+?QjVjaFN#MuCk>|c&9ZmCww3d)FSedaU?RsA5SD32mOY#w! zj}5nh6Qux7w_I>$$&zq4@zT`{b~|GT{Xzjfzrhy37H6e#(3Ph1LK~!jw{WeKhKP`3 zvm9uMj!J7H5hcUdOe(*p@^^*~&Do|gV1_L_PKjSf#K));|2?^=i9@L$+AD;>PVFG> zhHeVyR+FaajNp?mmagi{3ZBDAwL9%?4&j+#K+cik4*n-Y!AjqkpU{9)J3`KS%u+P$ z3^&1)1Y)s-MgP{`7obvuDq&Pn*UqK=YbO(jm~t3JvTPZ{9k zan3B)WH$Mjv(zY4t3}!1#dYG#l2heiV^i{Wo_3lxL`YpsUsJn4N&6WRcPu#3dQLv+8F`FKLa8(`+E1KYzAY&Ne8~VyX z(oQSLC1m6FU#J^sucAOE_3Cia^_Kx^3kZ%+7ijXhkZci}RQUvx%C+K8T^2o<%GzQP z4Vxm|NtlW_wge(|63rqE3+k-h0Z99D;ZOA+_`oxqA^C&)o1n_Mu9_Iu5{3ZW$gbBz zTzLsg>nr}HEPLL1ndBMgB?R1^KA+oO-NR*)J*y*%aQnMVkd%ouA|Gb0O2=B9B$Mmk zR~OEN1@xq^jm_UszH%TPxG>F#2xA&AGTQOwY?W2CHOL4cNA)+B?%=g{NHM^nFfh)% z7*zHcp(cs{APS!l783t9lC@^hAorj@QDib=(V{bQe&0+X+!pWQJFgfR{-;c7FNMJqM!vE_qOoNh;l1l~Om0&Z9hMC)`3 z6=X6h{h0e(@b+cq6cMX~$rT?Bt@BUWwX6mVF+n~*qJQ3Xhrxhp0`SM307K~PT!1E* z8Ht1Yb+)>9_gz?qe1~!@9!;LpTT!bvVu`fs4F`On0VUqs9K;J(U0ZpRh_f1~OZ5@~ z{8WV+yKt!^cjW&S9@1wdj}n3r^LR^qUr?9R4)dhi;A+ZFLtqCIn>v#p_;?>uZws6$}L{0GO zYq9{a_E%3W2pX0aFu%5W5Eunwz+u!oEbXM3a{oO3U?-|q9?l$26bfx#ObPSN2wV1` zxI|iSJ^`Xcb)7r!w%w0Ix$$GsDqIT8j;B~@jzXWN)!(Lgfkq+ zoS19DOw>`+LYo>>HCvTv@PD95;5xC085j@hhkGoJ3mwq_HIUuouI|cPW&%raO5YBq z8ldo@dOyNP#D7I0S);&%xAVlJDuY1{?%7z%lDm+{A03Fn_ws~P!VWI&Urco&v$zdm zkJ}I+j2=91Ssoyq1qlg6bjp%=rs=Y*L%*;7dyaPvJuFOPHaPA@PT6y!;QoLRx1BjV z0Q!9imqtXk)N=Muk+!i-5=!Jfe^(bPZ^e5G?ME{dd~;0f?OtW?hIFMb+PUfTVqC&4 zaBBnH-egA7$tRFVJ^O2D{mrjl&S4&jd=$g|qUhYR(4W-7zS3yHP7N)$B_=GE#xihj zo>{3ht)Unq6pS`R^KGLG2|tJ@58P1!9uOZnOd(hJN(xbLOLSn*N~qRMWkCw3rH}M` zw~05ro7~Bo&=G2rEEYaeZ*V`wFcrFnCzs5r2q{ChloH#G4!GA;V+wC7tcIdwSuYW8 z%ZWn+&1Qc$7V z^6?KiNcK@>)_(+>q4yV?t!}0jfB{)asicZ zT8@V8+9-&{CMwd-iP&y*0kNR&JW3FerzP0}eG@H_ki9BoNz%lN$e;ewfVU>p5t^hK zQdm0o-~&z0t3`(7|KetY?%Lab>sYPlw;?th+nz@H&3{8}I9eB)nT`1Sb*O}juCM-b z;p`;c-ZBX8ul~_Vq6jvQL;B-Gw{3a|C%^hF^o}_zkM$25@sK=?0QJFqGWMyY40!wx=lR z!7KjxCP4yqk0(&KsK7*R*d^3h%=a^s=cSZ&LyR&kM1Beu^e+Wo*sctf$Jw5M44(J`05bl+gk5Ud(sP5)S9Q_AE#SDm)S6=oNryqIfR@xO9td*h~pAM-YT|QF46mN zfh>tilgaJdKH?Wbe&`owToXJm(w}rQ=~X|96G>b?>MQLwdkH0}LX|44yBkaw)R#n5 zG9Sb6&WOlgynQ14P28_9>4rU;*Lz&f2H(|>@(&pA<`XX9J1dV}v|>Z-y0LsL>4kL> zDw%MKM5-AbUFb4gj6;F_)=L+(ntJC-ijQMRw3%dwnsJO_RukWh!eSXnW{2( zOsuxcitjp>Gb^L9lTZ6OnjWi+=`3Jn+bbZdVED`ZhahsT=dgE6MzAl*;OH6U&R>Sy zi;b0momk^OPQtP9X0#Tw2@kOy+G(OxU^`U8k?X~sAIOUB8g%HKm3jypv1aqoQFn?g zagl0Q$LR`3{y?xtGPsC`?n*`t5jz=3i|&ecXy@wl6VN3#%&aWl_@{;pu!qauIzwc0 zkA=|wP&-0#h1~}mz6RviyE?du5hq1AWtF^%6TcA5G_kvm1U%rm1`6rAeh%7wx^$n{? zn@Jje`X*HDA{|qh%={!yjI>db?O9C)VXxkFL~)D-V(*^YLAI9cao3m*a_^WA$olGH zV#B@^H&U#0<58&SMOSO2=6^LKx0SThLS!Jw2@#i5c#NN6uwHWy`4B#QdGJdG5=hH^gMZBJ<^Xg)7GTV@M z1+MKk$Y~&++H-nz0W;J@$-0J+wHgH#E0}D7eV9zoq&l_|SM^DaSacp;1QDEI3Xxvh z0_-!P(;q`cCmEUd^w5^BdEfcC?9f4|P60z?ttvM=Z`WfEjNR6++=S4(^g6DCfj>JYLrtP2h zRo4#u-F8lh_WkYA;uAU|7pocDrP51It>+bbjxf`c@X!bOp=_XkS`CUNNPlZRi#WC z`-`wwz9d9R)Px7sZbqU7#`NDnAI>Hn@_{%&pDHKNT_xc&6MvCVV{(Tetzs+ zL5H7dz|bBT3ILs8T&p38@qdZjeWK@4eOJeBBE=lX<%?|~Y6?GP|7+b%T|k=zEj%-v^{&@DN;g#ZtZq$~hssTG%SD8^qS(?+r0o9$F)<5QXK@ z4?mW$4FGA8U2`EFg{-6Gka>@ zcrdO9>-+3;9!04=pR0m~U@2&L^QrxXu_RX16{3(e-2Ly1KN0pEeId$Uqs1m4=8{wk&$!D8) z$yguzk>1;CqtCz?bway|27Qe4#GG=UT3jsk4J#!x-U2Hh&XaN%f2wa*Rf9%cpg6Hp z;TLHOplcc&(Eoz8_!27Ocx1Xzh}neo$~RT{)Z6*HB(dLQ-M1yoGn9<=&=2hRZ@1

@jR3}>$J5yhzZeW^h4tX=1KZv`@ecDu zR#)(@7)Fn2q8F6&Nc{1 zj%xx6w%6^eHTf6ByFERUXz%{D+F2Hw>J*d6&(N9_~R=`5tVLs0cDXXFykBRM30v&*`iDe@7RCk z+9fJNin`975Fc+q4la$2%4VIM8W$qwEfSv)?R7x?iU&Lf?r0^y226{d5iYp zA0Mj+-=4mRsCc}s!5@bTJJR$b_FfV_dY9r(UBbLWkY#eqmhVMv?Fx4}fj{fumF>Si zbllqg-UUG_(hI&5G>F^#`>C_e*5B~QF~S=qZzWH~b&dJWO#ISx} zgvZa?juR?B*X9bLY+~zE1_6&D>H(5ubs0(SHx~&aEDuYH?FEHLyWB4DK1IHpQ(%ux&|TU+zKpU&nDDslnq z(F308KYz&j9;h(rBKePqC`*>!tvx?FkeBQQA~H)DMGtmly7o&1C)(xbHn&}e=o~>- zK?;UDhgovKyCHe>=Iy=k?`C<6Tv7&^{d93TM5!lIZt0Ep+}lGk=bg|GkP5$xT*~ie zO516bUI%oxGi;1z*iNY=t;2!jD-$7e4B0zf*eaJ4xPb$BV?y#=hul0HpSw0bv zV;KdN_Yhi+f}p<1+WU%>*8-_V2;4zlAyVxumRKncsqc>O@aO$NOAPQ|8HX3@hJPXmQ=dNUBcXLGeb==Q){dQ5pYXW%;lC#*5pjMjv103 zP~6Zl%`CSPTV_U6+|8v-tpF`_wA?B)QLt{fg-e!-OB!l!_s~?g+tc&h^AF$md*AbJ z-+RvaI6VAb)89pycWvrW%VCyvYm6AG7z0MNF9x{5Y|kLxFFPu|qCNR^#ErA)L&2;j zsFT_~MaBM@h^G)IwOh)@HKpnI8wkD@oM=_}3f>5qRX^MSZrZFlEy%8@q9nW*o8Iog z3E96d$2gxbbgs+)uis#La3G{gB{-|$gGC$xcPTIAPu><0SYXGb0i!?3UAiye#J zgm&{Gjy(r-Zw1Pz_<&t&jjLm4+_Pw)C*SRC4w9xJGtgI9Dm&iJ$`(7Gzr%!k=D5KD zxf8sf$410>F`!+xA&wX=C4dyShd62zG{RVC3A5@tDZ5#>!)-abhB7w?sQI)8`fa|H zaTF=aG@fwpr{BOwoY$Sr8R#QuV!Y)(dE;1VzBNp0$TiDF>6VFV#w z5o60qNS)ph5mDQX%w#T(G_u_+*>t}D1bt*4>cOT5OD6-=noT2@ire|7sE6)gC7kcL zU&%cWYHy!gA^TSP^Ads*?ZyUJc?7+H8_l|RBRtvkJCGaB7puoSI_bNuAIEL{P@7Ni zoh!QQW!yvB19(4dT50kH`?5>-)$=pI_^Hz|f1y>~TjjP?=GEPy@6}wf4XO7gsO|jo z;mN@F1e0!=&vcZgPx4$D@QM+G2P@~?fK~G~r=SDYk9Kf|Wbe!w2TN6j%EEuQR za1;)!_3YT5r-XqEsMeY{*7Ytb9L_@5sG=8?9o2xPl*WEGJ-MqgW}A$gksx)iWJ|iF zQ*I|+9Eza<7`j0a5B&cNnXUz1Ob#oN)+G>uLP=kF{*#o}OW_F|TPa3ZFDLQ>r?_rd z?0$6rG+nSNjNC~9uS!9eLllG?ek@tr;tYf<)V(?+d?#0cp0C=xcjh9l@-ojT<#kI( zD7*hMPb|tl?O|)aX40Y}+Jt>WFWEnKwfUI;RLW!+7|pMB!k`Py@*f8W5CVuJ6;x`k2f z;iQ-j{Zc03;jsf@URH80z(rfz9gC68T&9>_`lQst8uVz7)}!5YLC z(i2$$_ya)NL|65Y?E=)LN25(^EY}T0*tS)Jkq|C()N|Jz6>~s4IKGFR9n->LHecP7 zvS$hkemfm+z9N?c%G9JW8BMpf@a6*Pm$<^RPA3X4XN$y4L%b^998k(fkn4CA_Vc;!N=*BE`=1Y_+D3tx)I~F0urOH!p1cC4JxWYiW zz50{V+2V|R$LgUIY|pbip?W%(Arvp&e%;A?Op!eSn(5Gzgj9c0H}SvOjT!;d@DGnq z&iVs?K(I?})ARVw#uDDFuai=ep0#nhL ztsWZ~`^)f;JuK=s7cg%#LW5=YM}a?*=z39)&#BW@3YhyNq`WP)PA|=S>ii3DBk;DG z$t%pt*}wp#UQA>uRtYdgXl;{eL(*kl%Z7x%4IX3roq` zc{=er`NA|aELuy?U8~ki(Inlfd$Gosf-T(qqz7LC>#s%D?VET|K*0tS`zoEE!lYE! zaU=?z%nB*@O1x8FGBhH#kPdAtk%Ua93<8@>HI?Puwvb*tyPPqumr}i5Uw>)EX66*F zpZ2k)kb>I>Z8!ZmL>&a~cZyZ8@+oQ#>@H8k}a9Qk{J+ ztmxMYcz>Tk!6o4X{Yw!TY3QV3d79ZJ40646p0V*bq@~AwT&1Z|xrK!bxa#x5ViMG@ zXhqxeTer9N+=O}V&va9tBK|uAS^Xh-2p9;-Xb}l&{4^N5+|)@!&(9pV<{nh*csOz+ zZ-_bs;Es7ECZ#M5M{JC}H~gBJ#K+HJ$?h8(fuWkm1sbKV^_qfoS8hIepq214;@O-e zg2uO}t}_-WBWT)PzF#koP{a|N}0SI>i`jj@?_In?-v^6M@SyOupwzjow@T(@REcor zz31`dNAh}-A8vNm`y>?FCGaaPE#Qr9uQWj~x#FEu8T_n*;|6wX4f55dft50gVhWmV zn}|-VxP9Sr#W1O!XZdwvad!H&j+YfI%a^hO)TX9t+&eCNmUnNR zd^>zi?%mb26)7hTVjLjco~;#iPW+{DJ*I8~xKm;L)+RtnZk!sw8lGtx#Gb!useJVBhk4g%Me7G=B)5fT zI*%_SfJOG=XEB|0e{6yxs@HH{Vq;svo+qg<&BZ` znbwvw<8K`-_SPSGACX^r(SsUrViGnc8Mu9Ypu{8f4G)a4YpyoOt-U???d?y68C<35 z_K`=Lz!S}dH$@-P)HX~myqQf^o{l5!Ggu9iyU^^DMx&mY`g3YgOe2?6q~||Z+E~Q5 zPnT4gNRTF$X|1&>ZFo3Bp|l6U&IKsd3@>D1dY?{~OApPW&*}?nw3_jK7fpC3gXRC$D zs<*`GRP0Aw-JtVodyH84{K6vXgXxRc6rCmF;qI?qQ7#sW}YT&yyCItMB{YzI?GB zSv3?r(h>WyzR8FKh#06olSzH%q7o5Li8!ecCU6A&3qh69XAmc1s&e)C0;4{|AZHc~ zX5i`e&=E*R!Z3|JB|6a+$Gq9k0uVcjwY=bstBzJ_WGDvJU- z8unZ0r`SaqAVi3@Pl%sL3I+>iN$?|!z=K4Pk&8gY!wlRAL@Qzl;S2DM8&M$txcRmo zAs|5*DGW4+kTp0&I2i)NiHhk8^AOQ{eO?NauV;#U{ZmVUfWexdno`-$F zICfF?h7kjZqzK}-{9>Z@_!7dW9L&u@N--0dr~C9jT4ghG| z`tv6i^zBCp{WdZR+3%Dr90oQB{ObWjqfkf`>=f)Ph5%C%9Wdfo41olb`X3BoVrU5V z-1nFfxLV(1Xt0#uVQ3SOx9vL&g$CQ}JIo06gL4sJ{$KP$$v%WYa8WlB6ag2hkmLo!UTgcl-1EW=xi(dAAvPA(f|Me literal 0 HcmV?d00001 diff --git a/figs/SmallSphere/SSseccal2.pdf b/figs/SmallSphere/SSseccal2.pdf new file mode 100644 index 0000000000000000000000000000000000000000..e1f62331c14349a8907cbb23325f4158f466257b GIT binary patch literal 19626 zcmX_|bzD@>_y6e*X^@nXPU!{#X#^#g?vURAD202&YYRux$`>Ddzm!fDsb|0y}@GoSy5Sq#Y@LS=WK3|B`%J|t?A}$>1ko* zM#m!KVrF4$MW>*lNvC7w=5FWgM90VVhKq+&TU|y|`<=!IQ98iQTW1T;PgYJIif(2u zwsscod~aE?xaHie%sc>tw`LwzbS!U0`FMCmc=>pNKLG(=0RvVnNl7d#C(GwP`Tl?3 zblj>|PBtF4bo{_2Ejn%mJ4X*74sHcUz{^`J3uj9!I>64|191Kc%P03B&oGI(#kKV4 zPItN_aKbJhzX@r-56jrJin@y|jE=m~(xQlT6>_-v6a*6r(+^}p6VuPr8WYkcqJ7*1fxb6sw()4|?hghEj_00ks|S-d@39|G){MOF&+t!e z-0pYQ3+|6P+wShXgHbjMPVQkFN29R&CYZ#-BNWzNc_q+kxTpQ&aPS37FHg{8`2Buh zdcg_%cGAW~C8O(72 zw?(84>nw6BAhMU-qt&>gmu&2_&B&N}nt`3aF2K7wha81Yf}d_Evalrelx}{sUuoJG zA#`4^S#Sqxln46Uu0f`1o_ZcOr=0VHt~XmdA7wTk>*{EBb=l5NEQ}P5=e8_hU;YLg zuvhC#L3YcAw3ZoRKQ6b$|EAI5%v$Yz2k-8;U~9jA%+o1CF` z>c|H0V%%xJHPStYYAj@n`!ylsb#eAqk85sZE;cIQweux z531Vr3bpzp`l}VP>8BkNl#KD*hOCzdom@W#1YXIsqdOB-!$*?)K8R+A|LDAReACSP zvL}PTPL82(Qv2$~xAWrj7qJyFXZ*t;M^=m7)Vh~)_C-0x$2;`=)ZG@~a&xoVm3*yu z`OFxT2CupLbOBu+>2=%@-}dkrrD3Ky#VqGnvxb>de(3egN#BHrzNOqepl$tL>Ezsu;?^0bR&4!uDUo3J_>y*_8cFs-MM99=6S2Vi? zr7iyJ0r}zvpb0NfnDpkkeKe7~Tg-u^u7r(YyoN@PpItK-iRjUI6-@oQWFxT7e1+tg zL!~9rD$j+z^mIWu!Xg(kBBS5K65H^-PS5-LZ5YU#G1FX~V4Rt?V3Gv>wiP*uO@`Z$ zP!(wSAy=c!+!ilH0ihAd2`E`N!`geisvMWoo0a90aISQh+OF zRA|3NQs?mu;1HOyeKU)Kd==yPQIge@rT^A+Zt*`4-$MlL%6!^4cR$ta7h?ozgt1z^ zlTOJJl^w#;rGRLK8AqW64rpatL}Adt-Xui!vnRO3c17;u5|n1X3JYIbk@6xM*$xs= zbN3=8C|vP+n`!>*N*;M7N>1FV8AGPNkx$^iMp_}9*tSxvaOxX z*IdBw3$9c?=J8yWl7B{@LxhSj%*H&QyMr@C=KL%338`det&WYu@GtG5;(z~4PT_mgJF}3C18plve3HTQ4_X`&3 z@0+c!x`~VFQ_;j8K|D3fD_B+~#BNipSnBhM#?Sp_vdLDK^|+jS>Sz;Jlw#P8MLu7p zSP7)X&L;5+ELS10rT>iKzcmDe@eZ$awO)ekI4;hUN`Qbxb|r^y3rNRNRf9WW5^GD_ z2dv|vj}fz8%%4T9_lquje78Y|F~7Py4Bqkye^jv=^UBwbLMge4@1$UFJ8rT!U_sT` zOQD}0_jYU)vHr{hwQMA7EF>(NpTL^?+dn(ul@{j!$Bb#EC`Y|uG!!!re@Jg$$-JL1 z)t#uT&k0%jO?i%^I`g9Dvb?JTqEx)3m!~*qhZKXH$yF-6b(?IXfp*|<2_nT-Y4qtWQc9b`xlgyL(tg6C@$2rmti9tpl)88O zz7NVyB_HqWMuR)d(Y7LN>Su75%*HU4qgO6?0u-{m;qgIcj_qO3efpVeKzQ9VgL}Z8 zM@@IRh38-WyzL_ziy`MBT*l?ifw#GZ9&pEEzOPivdVJetb30e3AXG|fJPe6+3W)oTelmGYW?)*I3<9}pYS-~6+WS`b_X ztca{z;f&&yG?{*@M_hmBsV;9@C3bwVn$qj}Y$0Q-((9bQBJssc5> zV{Q+Kr|1Gc<&ZV@ySWn_R1e9Yso6jw<7?^t+S8m>+h{RxI)mr0uIcB3c$qeTwfUp= z&cNi*$8EtWN%D1o&)nU_T9|Ajy?hHvwy}v(Z~IW=E}2_Z?#4GkOzXV(ouG?FaBOF2 zK5|5!_Fr*EI{ADxqGkJ&fU)>Dh&_oooCAf5@1tVS(K|RSI4@Nxi0B*#n3n0#j3u!~ zI8!+p;)H6?z@d?>6y`YcUIeFz_0`)nq|sc0wK~wSzdY9C7AeFn5tw*q4{Gzi&@f!+ zRJ`?j(Vy50HAuskfF)|&Wz-A8_0)0n@ryH2(;#HA8cID?)4b8_cfCKb0^EQ1bLv$% zb@%9~mJbutBw+-ir99;~HSCgBh94f)e$+&HEn$sU_0b262gALjLh$O&#M<|ei4B{p zf2krkGi%j|`a*EpmP5a1&Qz3HR}=hp>gS5I#vVEa7EX#H+^rUCOU@a3wfP&x;93)eY zp`TpOU~yWnqWKv}4HEvqi%aibY;o&JTJ|>j))MSP@d+^^k4^RaAZ}i~ zB?j;Df`=gm32_x~E@bS_k6s*-@d;tcN_F~pDA8~8R*n#P0>o89#=~=|$Z3k2JwrX3 zNb(9*#a5cHPFi#bxhT~FUip10pcgI@TYZlwQ!Q=2(|eTB%$vyGAFXufTv=JV?LW25 zwuxi7qr^Cvr4D{XR2@S1H@u?M5c!w4c6v;X)!_$^e;kj-AG7-9_Qx~z4`jDwt=Gp{g_0qLYqNPMnC#A#V{(2{cT) zeMd%K4ByRcdJzJN7{i3-Ez)%j;k6Hit;UZ2!n?l^=1ZYRD&%zg@~{yClN(D?8N+Fh zc4$)s(glu~W=6a-!tV7vSC+K4nOcz9FrMONxu~52>{Lqac^*=ha_=^nB8xZR9 zqP$!(S_|Yu5gxl3NiBI$$5IYMx82k4N4jZi#%@XF;5<$4+~NtV6r^N4Wn3C<-`M&6 z9AAGP`!0=O|DGJ?;SCbuV`i9_bCArSO7eliYsqD!(P`n(qR(Q%n)_y+Kl2Tuv8y$^ zGa#*ML($$1DuFb4H-8|*)vUr?DqN`JN3_)ghgrk~UA)QYIvBBpMn#7YU}Vb$xNxTX z7GE&lmp1W8Up{i(;d1haHs#wm;E~q>DnR_B_(6Jo*1`#XrF-@h zvzq4}rx!(C`LU_emM0z?23CIG^A;WP$k(vK3G+TWa*B<(f_UxDW+eH_gSEm*zy4>w z7zPN*=F4`N6S(Y471-)mEx%5p122bcMHj^I;00(P+jE%j*LKU!#(BM5#%(5PsM2!~ zMtMmD1|0TxR|XbFhtZCvfF*l`qCSUI^P~#+JQ~WhgN8F2*rgcsf_a=vx)NhvM!8kduG>;^Q>)E!nDFItgq=0lmDe>)Tz#bVkf%e zM|BsoW}lK*(%#zsrAy)^*PjY#}|NaPa4Wg88#$RRDSwp+fL~dbXMso9KwZ11JJ@jMKcypDxrep2+0cZVu!PnUn zNb2(q1abq9&ys5O*3&e2J)$E zWGd=?RqE<*aX+-F1A*`hsdK3Q2?Xfhuvr zv5N;>#S?hJ{`ZYXNw&|oUjAxVnT)(LT(CS{p8pTf;E{z;4e_=eB+H}!`qKm_(vBwb z`297t$A&@X(i^5uWh%qQah;2|a+n)kwMHNGDF&wsKkJBOU_}oBoAgg|_4}MijYTiS4l3QI59Jh`9h`p-LSY>HC1ZtRWLuIcwls z>7mkUh)|bx{sWbUTjsqqwoI*+PEtqthAPqyxi2OW&4;82yh2}DwmAQ{iKFmj>gS+; z-5krx*Lql%t{PaPuQ<3BJDF2dXHug{b)xA`l4m+!3UKqTT19_C7_6?0^UHwNQbqdr zdI^84Wx~1RTS7?YRBSd#JDFqaR8jepKY_wE@h)SH6fw=!HCoJC2F0Omaj^b1*A9+! zv#Jp0cX#hTvkgrXzPByEzlP^IqqTFWl3%e~0KI zkxm)O?h+n>Bu)r})}@#$#tORCbC6c@>k5_TjhI%P=T#D(dZyaWN$^FLIqlT%H`(8B z?rBa0n)+tvf~xidpKyPOLbe5yD~rE#r_P z1XOrt>$q>f3X15t3b<6aSs8Gy=x17`vg=udB357F*(X{7nF0i;2^Ls|b$M#=YR}ZU^1otPIlmTJ*uCr{! zs{RbGGayg0QJlEjhCqNgWOiN#uP$)CNGZynSaEuBJL>}kVUv%d+BT?;Pd1Um;T6V* z`Rag@Z3X3pu}^0Ap6B=ciU@Ia=YeW$5T3JfsUb><5LcwP@ta})D9A0+>)?Ksqw#H& zNi&eHDB;M!F80x7#R0p~$^k1Sko<5fMF~dgx?>;tP?d(DE6#!mMwUE^=uP4m_|q~U zOlC>{HviZX4Zxzy*%78*j2vju{1Av$zVIjQ!g{*`_<=-)gj%x&(5$b!BI#=XCWCr+ zpAEfFs*LK-WYne}%DfZitaEY-{X0YgvR8*HdLcR$pJ{zA$!SE**Hy2zu}7Nor4$}k z4h?*tQQdlk zrH8fp;}&(4w!MHZNtScZ^iPwIKe()NSR54|gl(r0jd)DrDG(p%rn*WNg$Dt952#y& z$Cr32IHIe_7oF44OL~KvK zX$i|a_9qjs0d)&}7^TocX>GZ*EDMEhs_*GI+MG+u1?r{KlV$F^C$}#9ZlXYX@C4RV zgjgZ?I5cf`Jo;(u!nZ?AB%BVfM3hzniF}M(!nEfsj#=2vY1g=&Z+n5xw~QqGD1A(C znU4PXrP*RX_oT*c_*W}x?O+wU{3f#!xMjFhX8xjTkS^ieSFd5Y5A2a7`|e{#)ywkw zYX{j)6n=o4sJmG)YQ=N`=G+P|+cU2Jdh)Uj`)MU%(+nLh`N8>|3yc1c%hWMV{JDoy z;;sMMoE*B{k;M6wN|d##+7~wZm%}?(c5kE3JZAw3o+B(qKF2}QMGNjV`xaQCoDidT$e-F1~`$%%%2;yUf0 z%>WsQMM930WmZzFECD^I1Gck;QbOh@X+(@4j?o`ZeYD_B8H7jf67cR1KAY7xgns^m z;AA5l5UAKKPTz1c9Bm!%{f@T@=*J6G*A&&|i(Qrvuk(@QFb}vIKifqIqb6||$*SPtkgqNyXsq&^K$CSmDDGu04_veESS%$y4?m6p#Z#p|J3_{$XKzDc;l_08 zLaT)HHnq^BG<0t_+jY)Nd_G!8kdq>X2R@Sful^_0PjcSF+JE`G24?um%NpS+JpGdw zr@L**%php72Bvb-NYG30PehiU^xa*jN#VmE(84jy@cfsK-g%Bo*9(%>zwLS{F7Vx| zer)>5o0P)y5eSLB9nTTWM=uw5S#GR?7TzWA zuYg*&A5c5?)?RoV5c0^+NB9j-l|*e!18aBu;?3iL>Vjy~VO*u>={UrvcIRdMZ+W1+ z(9U0V_s;Hc6qr(}_7P0w0o8T$FPtd(|6bzn)z<Fjp4Rjpv5l zZXf0%Iq#C;?SBtUjH>cLtutvtf!by1k@*Lyu39(|aRKStD-NXbekR_WeL(9pGml8` zM;D=W1k+?LetVvWtVtD>bH$uWnP9^WvIcgNAZZ*)TP1N?(v45-PU_?yW8(orL*hvl zi_vL5x8}}D{fs!G``FEeqSxy;wT01uY{>(+#b$<%*PV5j9vRlU_b4gS-s%a`ke+|i z2vjFO?~Qj;Q%e9=CA^h?mx^bVVwdo;3RpgZxn%`HK=DS28y@;9u}({Zd)s)t$tW5~ zPFziO(t=<%4yB z!X2m}yfYF^x-{j>s%acCozl^C>qsa$xp|_Y9IWI+9SAde#GF_k%iwB4Xq)InmvE-k zaXv3V3iP)-PIy#!(gD-t-R_J_weFv2$ObE2|7yE49DZV138WtWci9Bqj?nG6tAIBJ zUO5L}HKhkT+|cR{KP|u>?k1%S=aQpacepX7(vJW~CsS8kc)=DOys%8?5uq*0vjku6 zGjMR-K${Kq>aV5C*5I)c*ac=6#?tmW(XMDGd>mKF*%@=YNYY| z{~mIvpWvVF+|4>JiiE@-$C|dU=3!^8S4C1&*Y_>g?_lc^u>NhJo16ZC|7?=F?`Z_* zb4)I$%!BX^&$jD2PoWR{Kr(lM)37yL7;c38%Z#OgE*ATHsGW<2$q7u#G;M?y3f2r^pNZmEcS6j(|Kr;?D){)joOgLIZI(`*-0T>Kj!NZN*jVo<6>mM9Dl7^9&v8 zxcQhu+VIJRSTW_(j?(MG8c1uc+@DCYnSov39Ocbzzrm)^FkyRksQ-3>PtEj1EYZi~ zK3Py zA`kHTAIo*9t0Xo(paf9jQJEydMUrI$M&@?l4qt>m&W3g9h&pfCQrE-fI`AA%ijTpW zqJX44y|CpNxaA^L-N1MwT9vD_*Qy?9w#vYnD`d ztUUfhuHkfUS!ch94#?9wcJ-0LB#!5zgOpvBh{N9s&8|xHZie>NVN@4*qG% z)yV9y)`L>(J&74)zWr8fo?^!yIiOwrCPvWPt?jF?r+3%PP)l}M+OWozPEcdFZ9+y` zoR?BG<)Xa(qb2d5ItwLSF2$oZgZs!l^fT&kHT17{0m}jUh8X15dy4`0dJA7VwWw?4 z+PHF`ehc~@B#|Cw2v|%8oN|IQlp%Ij0O3%F@OtWJRehuA`>|p{SaCr`uJb{K=+DFb zr#2eAzeesY?k;^nZ_ZW7Ho2z8SPRmj5$=9tx$NAqos@6>rpKuGdQa^t{|?7Y3rL|T zDyXENwr3W(U!K<2<+waDZMF;qsx?@)yy}~I=kA&5S*m|PI&D`u{B=R3$*$qT2|2QSCpy}6? zV8$UD*F(_PN|Msl4Ln&<7Na@-SuOIS*I1QIEkI>Hls(sq@AL+2mO13h}Xw}ev{+Gbpt38}TMT52J{j7nB z(JSzO?Z>rwBj2N`k^htUBnGe!Eu;tcw&7da4p(j405}nW@FnA)9YwE0T-;|N0F6ka z0BhKWi4JMu#kKW zhRZw<7ucgcFlHAuWxI6`&0$NY?mdMWQKRz>VrXzF@3mIvOe4~@?-1J!Xt+W8X7+r} z4u4Of&0+cPOOVdq5YjV`A-p4UD%d z)O1#i0@J;cW6RYwa_q7RNy#&_&9&g|+Q%3k#!{607l;Yq5V9Lv{!jC)52vEDx^X;x zx%8{Sk5s8usNS#0nvC7+8M;&kyU_3L#afxOF(uAq+1%Pq?BqfKTp>jLeM?)3)l1O* zYoPWoA|Ha99gR7-GO1-)84^ewdgXucnFoSH0E!Aq9OFE1FeEl%?*U5Y%_|-*wn96Z zT%>5QGWuz1f@j2FpaL{jr{i9eVGv5T|35o}!G*jB#&bRBuEFUUJz+b{hZdRlt zPbh&|6b1c8?`GLa4EcWUWTZ;rxdHS)^Z*I8wi}&7sCD(`TlHs6XrfcDg*pLM?xbM9 z^5slJ$L$MR`FDl}&xpk53*x9r*_u8}@Fdn4JrAkne?cL92{^RTSFYlCz3YmbXAEzD z)kDnx_IM#fboTOxxbh);pGWNK6=SAW6bhHDP3jq?)p_cs)z&(m!n<+rkrELFvd{ik z>!QoAjJYeS{w2}7RNdYZUI1u7Av=MxROc5HKlqq;NY9IlIjM+93VLE&CvWwlYtA1D zX+bo)wLO^bV%Qh%BQFpQkfJQb2iU`jb?M_NU97&?DL!*0Arx1>sIM)X1r|8v$FV0f+_b!e^D0xSa$>x%Q5*rZnPG$~7KF7FfUlT;L)K#xJ=+=~RyNg{R<*qV?o7Z@=e zzT3`G*Q%TD_%9(LCtXYp?{&m8UXs$++Q%k)iL+w!zr`@}m|?;0D+Al3zQm;ON$fTU z2$jr7L+-L-T~E0Y(oh8?`9Xi}UaDgH=|h#l2ar>kRf34;D_IRuy7AXur2rN2S(7a_ zkb(^&bPiN{>s@|k4kg|=zK=rrWs`YsffOF7`(UntGwM3MqIW_s1~4A9`+3&>zSyD! z)D#J|W!gnePZJe0eB~eSGTkr5VUq^vhU98B=;d!kB;1$F!+^@3`#R75eb~EzblX)lxXaV35>C46RDFZEq7VYj^E4pXGf2Z0&JJ-RF z4$%#IWK(LI1-8)=6~W0yIGLX3ab3P#nMgWD48mkR4ddb5j5f}zZa~Z9-70ebP>G{WBe|PUC0P@k9GKXCM}TkAhCiISFjotVb1pN$ zYzsBLFI`ThDUlxl9lpM_cPt)BTV1!e^D-RJLn*N+#g!pt(j+B6oYirlUAmnC00|OZ zehzU~2+I$6PQ_uh$``R`;(%8uQ|RLoDq&-qGl_@0B@nmCpYu0aaCD{vxF%}mC*6MW z{WY~q0FuCGLO~XNM=`=4KQyz;t{X471Ly*I4(~6?^!hPkQJ|gQT8&7AUMC-Bxi~rI zeXBJ6@Qwj$(`5jJJN3&-->c(0IKBYZ00oE$_kq-m-zuh9A#1-rCzuvOjH87RhN9lj zv1d5y_bbHAaBIk4w@ca(q|(@japyIB1V!&<>#m4p{JLD#4ykJ&#+69Et#LyX?77yf z4TotT0lKJ9-th*ugXR0pZ?9=|he-o#IssO1E`ktK_e(@Hr*4Ms@Mo}7z1%ZhBi8XR z@@r}d6<_?)%+D`epo3=-MDZS8zo$iN>YCQPe6SG)r2%9Kutp-Umi~|_oN-%M)Qe;F z5NJ3qk;mdgBEO%2bWk9A+Z$Uwe4VkwEVF|u`9WPK*3^BizVa&NQdJ4Na6gqDOptN$ zo>V22UjMMjagiTr1{B74okvT{XWe zduG%h)UP_HT!TIAnIZ;gPB>K>#yE@dtfKze)83)mfGCIU8v4gnR_Qp4^1H#?I>%?p z50n(O!>MSEyEVZlVb~b|vkkA2O_zt6KQkDoj&qqMhIgQkbcFd?dunm3e4U%QOzNfj zBc-jkpwe?#bERku^Ls|IwL&JJU-5B)HSVU~M0sEJ6fj>5BLg}VLm%!#c|**aqTKO4 z&?Qlm^XxjL(=wRlk6+R!rEsh*U$^1`3hOqf=Sh1VefASIKSMqU6 zHz~Jc*4|B)Dcj8Fg)d5EqZHl6GvY#Z9OGt?ukqgE1SlD?AHN*ue&pHkMRmA73v;0MdGf%Dt|9Ee~55M7+ASOQhSxMs~G%-*1=lpYHG` zwE~eGLGuSf!)*mJob2#le52g6GzI5x8Rh3cEAsIYBX+ol(r&gOW*1S6gKCH7pwfc+ zcr++29AS~~36R#OL6LX)OyB8ez~W4Tq$QTf{&Eo_QS>e_QM1cUdo%mG*BcoFv||^V zTVvq10b6s)L1xFjzh=L?+B7y-W6(7^gP6ildZ^N>S1;OEM7AL2wAV}OianVTe-7hv zb+^)x$M9+MhvD}H@~#qSa-U)-mpx3YJAOCB{LP3$KcGBk8zX)DZhZW;pYfHXYj)CJ zf)no`cWy!}83_ryTsC#RNzL{9e>jPt$<5=}M8)(d_9Y`u6uy45Ofe@WF?7QMaLGAj zYE>Nh!DZC`&k<1SZ_y3^jN(k}P;bXn^L_OZuw@Y^Wm<_zneJ>Yx)^oGGkyxY!R5jd zSl$5My1JuYaW(ClmfHR*{b)@POE=sLifW}owl-&xO}#kF)!V>FODOI5AfqIhf00yy zH1ev~ed+s;P-4?+e%e2$!YOT{-s^aNBpIl+Jj96Z+Uny;$G^=g-Y-oPE)LCzlfNG- zDcVyAReFzEm_jX7@;+*DjXu4E)EZ7kN`hvuu#3ogQtNHl*z$7f)6$4Xn zgfQhBq41V%T)a~ERT#;ihe#{kk-Avy%>W6#)tb@NwOZHqq zR0ZX3ikQjASb34>)2e$3ekMOd|V&=c1jPZ>aSE63DOY^hREAhzU zqGT0RkRL4R2?e$r9HTAX@dk{#9Oku_X)Y25yqj-bTQPeSOtKZFWqx`}z<`Pl)?IvaK#9w~5!G-F+|$mhE_J{<4Z_p@`QnPAkA^b}{@$PRJ(qdmO*@Ne z<##hJH1o1C-U}Z+7MtN%L1&dUO5{-nTRqBV`Xx-UWAz5@CpiNvYGI2*D;DX>BiI-p zv&>zsHXCpMTGEnALs6nfQQQ;Ulnlz_|G1ldb4$7b`}5QUG`~x$Kv(V0r`P`4Lf&~g zE0#F%+wDRlYgp3#ZmLY$9r1FKtz;B${^$MqU+sf@iX*}lGOu`88fyn~Bio!Ofv{(! z;Tx?e-w<|dq4r-NZ_xEC|L{mrKE{$+17DCm-7f})rWL2ZkfIUkSr7GQdEEJeZ=UG^ zBLvTDRBl62Tik6d;!!oopGweN!J6+D7v>@nvi)ir_Z*csP@1r~k?OhF;CwZCLYU34 z@+Nm=$aAh(M6X%=Q&-WJ8EMfJYrHBw7m4UAHvImL`pCQgOI+q=;2(9X906yL3l zhR8Bez`7w;lz9|<3i*$ku<4gJlYVXgrkbMp!(nz7KrB>~!*D;z%u(sE?ANC24DdER z!xdAQVFq7Xg^6kc))BwX2p$0W&RrdK9D*N5>PX$gDPJ>JIGHkJ9G_$SynQ*7(P7{R z?K1Eric|Vof+JMS>&~}-KWT4Jb${dn-~)KHHBF~(e+}5IS`gcTS$ZuT1VT_eJ5wdr zUR@~`2Vxk7ACz|uS~*yvfudJ+&G5yZaSk&F$S6%mn_L$gku&cO;acdbhBLiGaQ(?HJMlzG{;dl;_B(sGnk#bl~J#TUbOZ%RK_S z%;|qhsf;CedgyB8Z0%f?-QpYofP+B}d&H2{BD=Vz&*nd@q#y27=WuAnOB$4*Xrot} zTJb5)iQwHfDj6=4G8xOUooUL9RTz1K>VLoj#OSI{ZH$tc?IkNurhQsNY!SsX-{7}> zg4R6DG^8x9fk?WN(=05Las?+v^e|v%6@jL&vv!cWo9rAi>=*vsjp=V5Ywx<;Y#>-( zjM8p5Slr~(f_(#Oth~o;M5o$yyZAI;xhsfP--pxJKyVQja1ui_tKp9tUgc@v`J)JA z)$l-MAigp2lle^EVO<~Tgg|pIs-ItB*9^7Vm zIwR61-yFs1JSl62A93SE^OHd}oj{*OD1<2p8I$vU2b9X= zgUxGYDowV}FI%MIPyh^|FrX@qY-$oCB#XKdn*B5x z?(QwCXTLT$_F-A!-Ah=d4;>?&tS*=r9XLV=mccJArrFwJ!X7u~TrjP`08CJ5;m6HW zu`yoij0>5+seQ-s+CzaIRGNb_*>XQ&kPr%~Ju?-TqaeL%b#pK6eFjV<2H!%8YJRCL zkD7pBTLr~TTvGr$p?Z?;%Yd!md^HkX5-p?Sq4wJqU=rSG5T3!TI4*sJRa-q>>)y#O z2n0|450S7a=2rfYGYqwIjy}RPN_sSTF7IQbX?wecxM4q<9r_M`yd{;I?9j^ zXP)Tzrv%TQ*rtavCXeFkv@|9c|2msQh({BXmPsN2DH_hOgMqVI%=^EI6FTBh+XCq; zx3I#4383yC-b)rkhOdI&k1oz;v-C@LE*1Zo;Z8#caJ2;=4zI~C*-HkXRD04PJDw}Q{>k&A1*HkYd#<8D+5M>QVi zP!F6(&>1;iUb!fL{Qc%RbNRQL=lf&BL`>u^75(Ugk5IobOhaT~{UyTiSY;4(L>gU7 zqm8-+0_Q(ABhuw0f>g7Q2K<{Tm$CIn_ZSjpJ~jeZuR3~r^yXKQk(Ycy0J%x%8wzupwf5ju+mz`W zxt)p?o-=#)vy>k^)QEuP%w$Kyn|?8l49eRa+1hY{qXnwdwbQt{qFms+W2E-Ssb)B; zMF*ux)D&1|riR8e_(+eDt+2(E(^bR#HM)Ugq;5Qw;-`T+EQs9Uy#7Q}fI^D2s1Utn z_KDk@B<~>J6@R6<No=4{OwmYo9YM^VUJ@yi*sM zzM2nsl}0ULlwbr(lcAS5cjWuYz0U|hn8evrPifpf0?tqFVM^cs00*Zw{tCVWUW=Qr z>xj!Aj}`%V;C1Hf7%b<%{$XU@*X>q`3lh~p`5U|G!u+q6js%f1XF@OMA>fRm6mjKC zG;Eu~r`wy7N^_eCJ z;z_FIL$~7)XZc+>B=`&;j4u44Vz#QyAVOPIgFQnORi%W!YI4aq11{rr)Tzzb0LkH! zzJE!G_cC*EX@Yu>z1S4_nNgT#jCA>0w-3i&Xu>)KPG{`R=A~O~1abD*e%9;EhrgZ7 zv8O!Gyd3^$u15L#;m`gs{Nn|ib)_gt?r!c7Khl|n1QBVE)Z&)b^;2zKZ z7nW^v#s`#|SI_B)x~h)$&`0AR@6K!cQ`-P^Mb$idjK*pKZtZBk4Em465D|K}8JlYM z`;PFSp1k!LGpP>)kUC=QL*{MJWKOyM5*!phYkCv4IkZmdWgzqmCy!Nf5IQDa zOAt8HH(Jy&@a$TH6-B>5Cd(#&+B!OR|9SM+H9$JnPMj zZ-&wSZ>PT4vHs4{c!W=0(=e_xt3Iq=vRPZ$&KKp-hd0z3Q@ z5S6HC_1d-0;Kx+n7G>ZFQoHMq&L%5kN}Tjw3dGYC-XfiAgI_V@t3EYi@uvN)`%h=` z1mSX9T{z55B-zG4vkB+9g??K0TK&=E2YSQ_{qr(o6}lcLIE`Kgm-nX8Zyeq{vlQg# zyI;5bEuWEwt2^Ln8jqF!QDT3}Wn_V#$zFW>lQ35B>9U#I zFZ@|kE8u>E@vsj0m#MQfRYz(zj_n94Rry-i)7I$8yV%p*6d3)?HB?foRgV2v)!!8X zrc)Qj@Z;P=b2jWQkZ~Ep$n2e=S~zUQZ){31e$a&-S(C+qs+k6r&y2YlsS9c2e(L@!VCtTiQ0p4>y2{}toM z;DFDbzStg%?gjwNp{v+bXHOCDuPRw#(cD^E2{an{)qpb78DL7c6;|zfZ+H%zk^ijm4HdhKY3;7AxaKdJu?%4z7tNP9D?-z zPd;z~?cvizS%gdyXW5VJ)qCIWmBuaW3;a47)Va~Y=nhrS+E!%?1T-;Sm>;dY&kwsV zBO9x8<{!sC5hYk z!gHkXV*eVn$1_OYIY3F+yfm~~@X;GTs}3t5QuiCU>#eV}c+2_qx+77NAt2L4;rHSM zW1zdlB7MO3U_as1k4;UrfvB_H1F1*+iKp|!0er5h8Lj^5e%?)^OXExX;U+9G z$dh|6Tpw26_msL{^ZQz;n6|#JDFfli;lx>CiSzMc5LVwF-0LWTcVTq9&{S-)H1?Rf z(R3Na&`Ft57EmL3`4>KS^-F5au(RRH&>#3P+Q<_~nEO>zZFdhcxUY)lhWQ-@9g9`M z3NTzRBuwz{c?0{JgaZ5_v`>^B4Lce|YQcED1woy@{t&6MssOShutuF?8-&PrTJS+6 z0CEI%4!SLV+U8;J_&>X&rzmI6+iU0R*-~fV!=Xd!X!{s9IJ5?HAJ}<%uOqC%d{L0hOaYS$k$)rxJV16!q37ln=~`k8)rZa7lxKUhi84IPUJcghu4m)U0#0ODWerQ zBZQt79uZ-22dh`3vDni3rX+a#45Q}-iK|kxyn*u@U+)lcw+wM!Q_)T@cVqKD=*M@< z2NdLHHSvnk(){i$az5EwO0UQi^JzUaa(e%}A?*pScrTi$D>Z{4q2^E~wjzkVH!y%6 zNDK2RlZ_oY;I6j|Tk#T^0|u3!IN7>aGQL=T@D6dPRl?T$j9fzG4IA=~Q=7JMMy|Wh z$d#)A4XYnn-iTyg7LlQ9rYY@Npty6g^0l}+m{JuW|J$P!ZzY2-gno;keF5GuAYq)? z6ZD3RF9|dm(}m71KvlwgpDPc2004^EqxV(b*O?keXH$R_8T3Hb=1LuWv4Z|6v;liO zfIVI6c3uyc!NebqhrNRzAQ|p7Q+8q}uSB-{kk5>UemB#Ukx23HOt3eT_ANC@)O;R= zlISW24_;w%r4?uF6A>-N#%b}Y_!viCCm#{Y^t{wh;^jRS$@3xLot9B-(uq(rh@q-)56H|=(Rfj4D9g` z`hC*>DX)xKaDqO7uLq@k;;s_S}qJIGFf#1-83Jg;G~uxjA~ z=`7~T3Ux#+24W3D=v?OMAzR6u*r-N4%M2 z>Hg})cq?Ue_2waqcULbygc74=;RN&i;%56M%K9+NKP z2{ylpK^1H`1dE$g=20S!QzG zV}$b_Yd^#m!W3jpS{aA*@I#VT#!YzhOvW99!!%#YOql;mnShioWRCmVpDm#L7kZZU z+E!SvZJln+MHa?e7_fS+(iA#}?NKXrm#kBQj5gJ2N)|(hPba2pa5i+;^4H#dLZHcUF&aM%**{5+4&8I8^r z79BcEqbYFXTwDpK&)Bc<_j?d%)%0pL5Mf*wD5M}rO>YZ7XXw*F&jLz3V9emAZJ0#3 z1IO}A*Rw5G0hQt6Ei;+mj(DUEESeh2Z8a<@%#zZ~fy&>u-3 zzN47xM5z={Bi$VEzQeeVc3HIc=dk#d6`V%;r=r;Vlpc@A!OghVR7JDz_)YSGifph-cN8utADC1*f`i|% zLj^xe4mCudCpVBH#eg~iHfeW6*J3WieOrsOnBf9qW>w5^@D&-A#6<~^hpNwwKIadd z`$G6YM&%EzA-&MUj)5iYSG<0{ZmG7+XRNaAXnrRpl_syfJ$7?DQfglh^wxog-b42SJ*Ev;y5bRc))l%+5e$WANYAvwB{OYqgp9dMAtkc`N!HFf UG=fLi%@$BK?wO9ODL~b}0jT_<@&Et; literal 0 HcmV?d00001 diff --git a/figs/SmallSphere/ScreenShieldingAnalysis.pdf b/figs/SmallSphere/ScreenShieldingAnalysis.pdf new file mode 100644 index 0000000000000000000000000000000000000000..e7e3d3be358ba25e8da6bae06982d4857e30983a GIT binary patch literal 9369 zcmaiaby!qg*FGXSw4~&qA|Ww!gD`-UbUM@kL&K0VbcZ6KNGaVQjdVAHN{Euu!T^%e zAn-fm`#jI@{l53TuJ8Oa*I9e-weI`g>&)4EpT(x7pv1$^D+FSDS6WsM;s-;(NJ|Hh zloW_h%N=R$Wrc7DbI7^Et?UqBB~2}`0m9wG9_a!W;1%M9@aSsDY3V9!>PdhBHws8A zFK2{{=OcHxtDU`-hkybnh!5(HfO`T16yTl+Fo%MK0EAzhUw|LD1R%l!PdGs`G9ZMD zH72IOKe54l>IfHGPdl(6VA28eDcL)D0yy}ToB%Hh2rHyD0t}4v@C2MYgHY+)8K2|W zn%p8zj;Ort`6M3WT{qx4-NbHvDIH83YLLLzqs8!2w0>b84M8Cdy#B6N|%b9=UELjkiGTZh3jZeG_y?#LiN0j@$*Oz2j?I&a#Yb)DP@eK>V zV=rcBkx1F!o2x8=>k>hn@4kJNR6lBJV;vt@Nt%LAga$k`v|doIj*uSQCmCdmoF|hX zu3;CiWFxZP%kZ1rglV1<4IYsU4z?9+vfE$PmCaKQy+TWOJ@QB0QfQ|dkZrEIzEm}m zGO5egx4ZpZu;7g5H`?3xjk70h8jP-cOog|0RcDpsw~4B~xsHru3juoemEEvutof2z zgK5|SCrd4q$KT7NGOZ${ZqGjd9Q(L=BUfXoeUox0vE*@IfN#mN>eUQ+XG{pEn3a<( zC0PU^Qo*=FQP3 zTMQz{2Sr*P0+v~AC6pq^VUi}-kA7|6#G9M_X|5b2%6C>uds3Bn?cKe@OSlJm=qlS! z6Lb~OxF7H#WdUDi+vqtFCt>kEfez$WuZ!DMGLE6li~8hz17u(d!>5w(U5SjQ6~?pJ zE_F!{UJ6!+MZ=q(eNVG2zDh_-f2MEAL6Om3KbENG`kA?C_%7NVU$BWrrtX^oJg~w- zW0ka!3gu+qiso>wbb8+&lKzZIftHB4bNw@a0S@0G=!af~0RK6fzQ;7 zn@F`cmfu7gQ`?g)1aBW&?Kw6CiEC!v!ee@JmOh?W?v83rlwl)oPF(kIlBGS2(5Z|V z!3x{%EArjNt6tijHX9SIw#7nR(6H4p?go%Ac()#Bj1Y#I^{S|>NNT#4hjtuS1NWpU z_lub0@_5vC+-3?6IpL!L!Jpmopq=Bv{N{8u$Q(^VRc~r50X%N!FoQwArbHQbV&CT# zRF&qIhx~r#e1)2T3qh@WKVfp{GR65V&r>7OMZr#s6%URozZt4gIaS|-;gdz(?u?zZ zgB~~>R_?-|5I`BAMCX|TCmKSr>j=K(Z9m~$(ili(4ruS7YNRD%*-Klb#G2l)<54`% z%i~;zzCe13drpnJ#z5AE2a=!McAg%ndE2#aKGRSP=_96A{iPGo;P&(d700c#Qybzy3bd>M~YQ5c1pwsLf+W(iijA2fo~rnr-z+#$A}n2&?6 z(4yTKV*xXF!F_6}v%;7d+H~aPXSCtZZ*z&^xFj%Z7U%!gwS(jpW7jney)=?k3429c z&+l9Qa9Z0!^)@K+M<@e@gkG5JTg_>l)I%l`Oi+hvYVT{K2kJ&f@30)pxWKBjS{Q$w zx0A*E6Y2OpdshcsC$S0~XcD&iU#=j8x4a^7nb&RC!!f{)}6kc_dJ8qsNbfbaEP* zTEqs3sX^yT>d-cN#e12)AhsR42yvU@m-~_AwDI)`F_K5*@3!b7IDzndrL~YP#R#>K z!1rIUtUM;O#iz$G5C}*N7-ocbbx2Aze^rDoj9aK`lL*-E&!rYUggxwRH+sDReeW>( zG`#@uizWG4JS@t$ZX!t=RSdQ-;v43NMnv1;rygAtfjvqiaiS!xy6 zdj@17VS@TEGZqKCiK*ok)Mt0dIKWk~PfpN?`|r9l^B@3c#;d1oEnlg_qF~LRJ;|Fv z)=ogG6oX83-Mm*9AFmZW&JG%-o)RDw?9U!Z%q8D zRn25&TV`Kfh)w-2P9x_BQ<~5~G3Z8^lD&Zc;2oaEdYzEm68Y~jnQOpQf8uYFFlkDE z*N8Y&Kxda$p^HE9LveR;dyY{moN$f@52)9fv?6Ln#Xj-B3w4bwt@X2B*?()$qZ@NB zoUOL?f<$*n3#-S#=TiepA%%^{hH{-zE2I@IG{X5VWZ**tt4Ep5R~;PQUiq6WwR|)c z!dZs!I3Sa5hJSA7E5>=|!e3N4d_%w3&#H$pj|-e2_vo`#^7Nyr?bac4LdB?}AznAL z2=Pd+-n`Y9{xnA$O%oa2mbY^%zv+hcN36W3H=_C$>?XromlhqaV`wJtt@(ZX_clXt z*D+xr8YSM|NI&%4)HivL#t%>HX-H8*!VCu;QCD!?&|Z#i?La1WVg@`uUrCsh*MQG> zYj7*9^f+x#TyA6iIsY2-DivG zSoS3&);ONA$^Ni2x!kd7(f|#u=KS^-g(J9;wcYwg>Wj}K6(tV=m&~LNj}^1jNL`66 zpQOUZqV1w&*g58!be>Rr<5u~cVDK7|J2nao2TQx&sqRg1MD{4Bx;vGgw5v4`sPf&Q z(V`5%HrZs9{#Ls|AGJM-#wjQ=*gM-$3n9ZQ663}@M+R%&80uHX&nl=a#!+be%w}Bv z*1yLii!zbdHI!}s;9#HUUdNNBb36=gzCB;7(G0Rv8tInah=nvouqo*bILmJ$vJ)&? zxX;5gS6E&RL~Tkcd<}JXl`B;+o_=d8=}3_n$H(7k{Pi&yRay|)H^n7E)^sxqJCR91 z%50K$FFeGb*JV{dy5hxFCxB1T>YhFE4q0#J^s_RZeAsc?p?HXAd>4%slg#xz9s>@v zR(;e)tPTqB8zHP@6YMA;qR2`^7WR?GDR&36O-Lq zO|THbV-shh5zE=A27>&%^uGw~$#oCwiFX5~=VDxHC(%5iU#y_6uY8wrc$+WdKS~n> zR!$n@6kIItlblclYYTMO)c@j^Vy``3_rPASX40pZchHc)Gi!_vGUO*tUt(%j(ehOY z?yHw8NB<`r!oYcBCJ>@Uq0(Ml*rZC{4a2cFMXeJh85WER12 z3Lf=RVVB0&yWe}!jH-RmJiNLK%l~T0iS6JQgYna}t9$^J6H4?@TwbPz zOxUks=xX}VfjhXSiKy2xs8`Yc4UK(LNPSNR5xbaI#=WNnjqMbIU!SbaW#GfQ3D?5k zWWQ3|X3bSw+gkYGBtupo{y?nxlS# zfjpC%yTdW3lJIarlmJg;#2NSFS7%`>;2{X{RBA%;aCT+9bRUg%w|~=F=9mMBb;I+; zN{*7@dIR24GDv~f1+bR}7j4`RIUKr5SyZo%^~k5QK)eyJC`AIjLp)u&)uUO&st9v! z>3pYJj@kq{*MNvG$Kq0%^0N2zJ==Nz8;v;KXIXp8CXU#r2C{zo)08;v#S4jYm<^fKevdtcVUFF?tqe zp4IV>vG*gnK7^2p+o(%@v{}3nQBSY2do? z)A}XG9w|{msTU_zc~8}cNX$99iK5WDDfJJ1Wj<-swPoseb!*}JSJZX9i5AhYXYJOA znI6l0b~l1(kD=r_;IbFW*ec^vp8^^u6qQZv`@*izA*_SjFJ!A<1<{kELs&+ai%&)!ik*vH({SURWkS8|zkBc$FzRGRCMHuOHMs=427*!h3c+l(Rj4-jd-F-9y zwDB$q6Vp#GkW(-VgH@l!%n4hc6xd&`l!0g>xd^<)e)KZ2$eUMmIW_b#x_1QdPB6_) z%an#jQ+dxpM`DRhoosWfG11RxSe;Oy`?&OtcxlEI`C}F;=E~)CgA??JjCJLD za)8cUyiHvpbG?Es)F<_E5f=>vBqILQujNTB3hsGKRdD|by{=`=AN)9+dj5%*`0z}~ z9T>9?l@Lq3&H6Cx(j8@B&%}Oz8&va()6TLO`}M%BZcZ_ZcGkj()u9@p7bSe}7!x%7 zc4)%=Z5+h!pT7@O3%*Ikih5Nth2LKt97u!Y|3J1*MJGj}Fz3IdluvJB_NnT1W&2yF z#u#8O3A_=@(>Se}CgMqqL%tHAFopvgFT_5egOa{7MO*0cjX2?xIpvWU(~{133+}xP z;N_}KjbnQ6{^?^AuK~kM2jdY2s!PVp$UI9+JzN%sn782!?7T3Wh<5#@TO=XKw|pqG zFmZ$wPA@dp>y_BZlIpv1Zk>tZkhrnH0SJsvpVj@5NF+(5FG|1Stggez&)_CHdpPZ0 zST4wQgQwbzIW(q+@XjaE^zfNU*{4rWQWN2nMv=W;y`$w$gpn;ttCLzsj;_L4K}0k* zseR#m1-SLB?ZNEm1BvI((jMN9MQ@&sy<*i<5cp2^F@&nUt<>K526jNxh+Si>(R}av zX{z)=G@Mtce+_SN)z9u+c+2UkmSeL(EbxB%kG{?~P_r_Lh2Lvy3nhn3neg=GwE@@) zP1U{AUK*C*h3_|zH@=FKR1Jm1=tLK~iB>1|$#>L<4h0%vk&ux~3QU)Lcd%;KhGNy& zAbW9ir|&`%URw2|!`A1hCGcsr(Yb^?-}Q`}0+tOZWm@SMvw6E;GQDHHV2(jQ@eh;# zG}}j@SJ`X~tsf3;Zc=xG&N} zV<<(?*kZSA@XN#?#!S)(80)XJGp=6Y7`{{o=~<*NV}+spjX&P08#Qkk%^b8mSt6ZC zkhb~FZwHN?@Pmc0)TK|(bkWy%JsYUzqH8Osk6&xj{+-K<7@W`-<~66qi%W7m;~(>B z$8-EJaReJK02$I*bo3hH@lU1_snu7EjI%{=IUOl?y;uDqEImYiS#4Voz9}Dm&`pus zWw{({auCfk%_4jjL$u~9=Ry!G&LPOU3ZS}J`QVDUT`U7KJTKqQT^b<2)E4~-(SEqd zE7X`-^)2M;3!5HPq&kAS(XX7cn~%JG&~Uy>L`bn6NeF$5p%2Os(jBlSTD2 z_clL_hRJfEJNHpc^pMe&SNayUJ$2zXO{CDgd(KDoGL^CWuGKAe_)Z{px>pVk)ct~U zVmWWZ#EcRgO^9^Gy3#q-N?gm|XscPzMU^twai%X0QT+z}G)78W`*;*s;yV%IH@Y|7 zFt0}vMBT2%8HCqq)pgYYFxV6Znn5p&EiB)fSG?KP0um!4Uf0;`nP5WZo75Z7KFgp( znu%K)kWWvamvpsReO;y(yyDx`Pj<3|H})m?QuN(1{Balh z#ps2_vwcQ`CUFJ}0Nr=Wv@i)r{?-kZ=|Z85ek&+ixhF+b_>8qMI%X!LSr?A)g7u+4dUa1{@pzuD ztLyWCt}z|u%$0PbBAAI;CUJvW@Ad~)y{Dh~xO7=+BI7EQ^K0uV4g9ZJsnb!t~T{x-)&gGK&GnEwn&`%+DvvD&X+d| zSOeuXvhGK`lR7?k5>;$Be)9lt`3#moPm`vJZb76z#URWY+ML-qvA1eiSjfNj0yi(- zT)*J6dj4z(3}+NNqSv=Lbl2dPheYlo%;>xKSnK?_=j(?OCrpoc$Gh$B;6+TDOccBT zsYP#%@?dYgV!2~E7d$ZNaSX9*-lXxIxH+2(1J-8^Tg>nZ+NWJXtq1Q6$gAY zpd)@5Ye%_p#^Cmla*Kot+eBiVtdDtYw)pDH zn}$56h1VYrNP%_CaB~XjmFFaPm##{*>UL`JxZkZ4dw^kE6^G|ef4p&eK`!5vAQ#j8 z!7Gg(60E5bWDDZgomu*LkbAV|G}vh%w=;3Ny$^!sh@)z_G1 z(mc_XDh+DR4g5&*qKdvWJI0a1(b8UxJDakEqJ;TG9hon*Y5QXw zm-gJr$9QEM%;eD4Ir%|XsKyz6CHv0rOet`?2n5D6Z) z)|DCyp=KRDLk)4-*22mI-I^nJC9yU!YG%@PblENHqt2@RoO3E2XIsmq7f7p5i-9dvRoK0jBwk~Jd73Z{qk^wE_)tkJ!-{~*&FH;*DcoF}W?)#%W zX`U`c^TB#;rcc~HuaaNi{DOCSdL#_1yQy=u!E)apjcrcrAR_s>sHe_u@DIj)u_&`o z1uRBCEy<0`7A^*oi^hRxwyPK(r0{u=wVhN@NB?u-fT@}!uo0NLLFvkxDP97Arw|ptp)yrcfI9t9q zxrPdgN`7^r@w}5burZ~-dfb`p8cNjM7#P`qRjzmxrKZvtSQ}GlXuE1|B3V{9qwsno6 z;+|9swdM86w6O7~O}f*)?m98-Ok%C!Ut`AatkuzWoHQo8;tIkbDwRL6qNd(spSX*_ ziKJ*XTo+x+CnoorLc7#=-X=3214>p{@FpF;FPZLdU7_FpY|*j^aQAV=*X-_pkYeV? z2yHNBvx>(ke*z1GbtPWd%6w#lN;aQfWM~T0|C^0auGZmTtNqu*}|S zt@_qu zwbG2T8lp@hLrCsgcmMnZwxMS7I0Xztt5BSgj7H{BD%3H@bV8qo28UyT|>^1(PG({-IOx zRyLGZhT~kv9PL%xyK+Zr0f$^)ux2-@Y|*32%3!u{y*2>eAo%29OhHmGRIU8ttMUQELlo)cD|8<|3iBwy`?loGqa&NP_g>kIX{G>S)Uc?FHSOMLwjCm4b9{NHF-efNAKqBBL<1jPnq?dC zO?O%eoc-4IwPxd^F&4sOqy7GV+lCvwXTWEa0dtDDETtZ}`S>J@wxxD6GdANR%a7jf zUSnAQh=@=6Ksz+^gQ38QUv;e>(${2a%k;vA0#CXQ=ULu%Abi}r%{C_caNV6y-vQ}o z!u=FyCkMTZPd~=xt?Dp)**y;`*^q@{zIwp8O5(2e03n)@mN5VaX}5D3iAz!p?bq>7 z*FCzktzLB5R*tL!G`*v!0jJ#Gj&vbY$BjnFkhwg}hn4&v?JnK*7ZHyayexr}Gh=<; zqpuvb(!5U1HhaKPRJfe2m$OUurt?V*?%Xae8F09jRAB2r&TPD|yZ=%5#!1wnIDGW( zNtRy#Z#+@jaF-9DC9!X8L3OPTvXp=Kw+Ydr zV@%-5Se-v)96k7bnJzHc1+VJYb+I>=g2Tq2_Y42pu#{UkT;=Z0#!uCdB64mqpE{f) zqz68$w5?I<7EL|vFd9Xq0Z}Cc4_dANG@EKp+#&W=6nfQ6k*rBebZ*n?LV?Frt@q5n zc8-tcuxGO2R`tm!ZMoaGs5RUjR5U!_AbEjhZvb)gn z-8M~tD&66_QVqrI4sFVpqHsy3od>l>>_DFBrtiaU!ITy!A(?wk6qF+KTBiV zwgaoPk;96x@Dt?)lFT6O!@WthU9an`!(BPembh=PuVS*18dNeHn&B;#g_ris`g)X-?OySFKN$Y0!k51M$EG2a=)K!o-7BXTy2^%AZVFs zQy}c_-scZQds)hoccnB5<0FzB3KWv5&N58xr{q|JCVS^(_LSygaZWF}eS4<9F*gsl+MIh5Qc@oCg zul6zSk;|{%gqzo;u~FTv;}`H8c0#s;)GZ?)=I0aEfmtUju#HtSUDePF`?F$Wcb|?m zaHmN_ax34;k@6>*zLS|bdD+#MtB)hHG$kyiQSMV0txTsRkB$=DC}Z2ov$R>BA6TNe zWZ%Qijy_)BjNT_^!c-mnRY34p{Q;)f08@t`CLkp8cTIxOUp5$~I6=qsZ?o`UW}r+0 z4Cd28cp$yptq>ky5lk0p` zE6l&^;Q`cyAiV&0{2)FxdutCc3?rrv*c4OI@z+d<|1%O3>t9+z;hu0Oq%FpPk=5h> z^HhfeI5Oho)Y$BGF9ss8s^Pj|Sz6ELZNr7{xf z3C76(CyP~FY>;3~NzlLU6u>YkK_PJg2vl5NTuew&PzfR@FDD=`q$n(+z^^C_lmLm! lfdB6bkD2R